[wip] fixing context generator for any_of with range

[djarecka]

fixies #1897

I've tried to fix the context generator, but I'm actually not sure if my approach is correct. Should slot_def["@type"] = "@id" be added if any of the ranges from any_of are in self.schema.classes?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (1b6ca53) 80.13% compared to head (dd9567d) 80.13%.
Report is 11 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1898   +/-   ##
=======================================
  Coverage   80.13%   80.13%           
=======================================
  Files         100      100           
  Lines       11476    11504   +28     
  Branches     2956     2904   -52     
=======================================
+ Hits         9196     9219   +23     
+ Misses       1734     1732    -2     
- Partials      546      553    +7     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[cmungall]

does this definitely work? All metamodel elements should be accessed as objects, e.g any_of_el.range.

[djarecka]

it worked, but I will change it to any_of_el.range

[cmungall]

As a next step for this PR, I recommend taking a look at:

linkml/tests/test_compliance/test_boolean_slot_compliance.py

Lines 54 to 116 in b6c62ba

	def test_slot_any_of(framework, data_name, value, is_valid, use_any_type, use_default_range):
	"""
	Tests behavior of any_of at the slot level.
	This test creates a test schema with a slot S1 whose values
	are either an integer or an inlined instance of class D.
	any_of is a special case of boolean metaslot,
	as it cleanly maps to many language constructs (e.g. Union in Python),
	and it is often possible to normalize complex combinations to any_of form.
	TODO: resolve issues around combining any_of with default ranges or
	asserted ranges, see https://github.com/linkml/linkml/issues/1483
	:param framework: generator to test
	:param data_name: unique identifier for the test data instance
	:param value: data value to use in the instance
	:param is_valid: whether the test instance is expected to be valid
	:param use_any_type: if True, assert additional range using linkml:Any
	:param use_default_range: if True, the default range will be included in addition to any_of.
	:return:
	"""
	expected_json_schema = {"s1": {"anyOf": [{"$ref": "#/$defs/D"}, {"type": "integer"}]}}
	if use_default_range and not use_any_type:
	# TODO: undesired behavior, see https://github.com/linkml/linkml/issues/1483
	expected_json_schema["s1"]["type"] = "string"
	if use_any_type:
	# TODO: undesired behavior, see https://github.com/linkml/linkml/issues/1483
	expected_json_schema["s1"]["$ref"] = "#/$defs/Any"
	classes = {
	CLASS_D: {
	"attributes": {
	SLOT_S2: {
	"range": "string",
	},
	},
	},
	CLASS_C: {
	"attributes": {
	SLOT_S1: {
	"any_of": [
	{
	"range": CLASS_D,
	},
	{"range": "integer"},
	],
	"_mappings": {
	PYDANTIC: f"{SLOT_S1}: Optional[Union[D, int]]",
	JSON_SCHEMA: expected_json_schema,
	},
	},
	},
	},
	}
	if use_default_range:
	default_range = "string"
	else:
	default_range = None
	if use_any_type:
	classes[CLASS_ANY] = {
	"class_uri": "linkml:Any",
	}
	classes[CLASS_C]["attributes"][SLOT_S1]["range"] = CLASS_ANY

and adding something here:

linkml/tests/test_compliance/test_boolean_slot_compliance.py

Lines 100 to 102 in b6c62ba

	"_mappings": {
	PYDANTIC: f"{SLOT_S1}: Optional[Union[D, int]]",
	JSON_SCHEMA: expected_json_schema,

to test the jsonld context output

But let us know if you have issues or don't have time for that and we can take on the PR, thanks!

[djarecka]

I've added a check for JOSNOLD_CONTEXT. I believe the result should be the same in any combination of use_default_range and use_any_type, am I right?

FYI: I must say that it was a bit confusing to me when pytest was failing only for one combination of data_name,value,is_valid, it took me some time to realize that the checks from _mappings are only checked for one combination and for the following tests this part is skipped

[djarecka]

@cmungall - I've been using this branch for one of my project, and I have problems when I have any_of with range either str or some class, e.g.:

    any_of:
    - range: string
    - range: langString

In this specific case, langString is langString: class_uri: rdf:langString (but I believe it could be anything).

If in the context I have @type: @id and the data is a simple string, I got an error: Message: Node <http://localhost:8000/examples/activities/items/item1> does not conform to one or more shapes in [ sh:datatype xsd:string ] , [ sh:datatype rdf:langString ]

Not sure if this is a common case to have range either string or class, and if there is a way that this PR should be changed?