Dtype numpy

[evalott100]

Replacing #315

[evalott100]

@coretl @danielballan

Despite the fact pydantic will fail correctly for the list field here:

event-model/event_model/tests/test_auth.py

Lines 78 to 88 in 94c7801

	@pytest.mark.parametrize(
	"dtype_numpy", [
	"Z",
	"i",
	"i4",
	"4i",
	"i>4",
	">i",
	[("some_str", "Z")],
	]
	)

The schema will not, it's because for some reason pydantic is outputting prefixItems instead of just items here:

event-model/event_model/schemas/event_descriptor.json

Line 76 in 94c7801

"prefixItems": [

[evalott100]

Inserting "items": false makes everything work, I'm not too sure why pydantic isn't doing this already.

                {
                            "items": {
                                "maxItems": 2,
                                "minItems": 2,
                                "prefixItems": [
                                    {
                                        "type": "string"
                                    },
                                    {
                                        "description": "A numpy dtype e.g `<U9`, `<f16`",
                                        "pattern": "[|<>][tbiufcmMOSUV][0-9]+",
                                        "type": "string"
                                    }
                                ],
                                "items": false,
                                "type": "array"
                            },
                            "type": "array"

[evalott100]

For now we'll get around this by adding

_ConstrainedDtypeNpStructure = Annotated[
    Tuple[str, _ConstrainedDtype], Field(json_schema_extra={"items": False})
]

But I'll make an issue on pydantic.

[evalott100]

Okay I figured it out, we simply can't pass a tuple into the jsonschema... While it will do positional args differently, it requires the array to be a list... But then paradoxically to generate the correct schema which does have different constraints on prefixItems pydantic will need the type hint to be Tuple.

[evalott100]

@tacaswell implemented this to allow for tuples, but it isn't working in this case

python-jsonschema/jsonschema#148 (comment)

[evalott100]

Upgraded the validator and all worked as it should.

[coretl]

fixes #308