Unable to create FeatureSets from features like specified in the docstring

[Zethson]

Add a description

We have examples such as the below (slightly modified) in the docstring of ln.Featureset.from_values. However, they currently do not work because of https://github.com/laminlabs/lamin-utils/blob/main/lamin_utils/_inspect.py#L176 which identifies all passed Feature as unknown. It should check the Feature.dtype instead.

Code:

features = [ln.Feature(name=feat, dtype="str") for feat in ["feat1", "feat2"]]
ln.save(features)
feature_set = ln.FeatureSet.from_values(features)

Error:

{
	"name": "TypeError",
	"message": "Type mismatch: identifiers are 'unknown' but field_values are 'str/categorical'.",
	"stack": "---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[3], line 3
      1 features = [ln.Feature(name=feat, dtype=\"str\") for feat in [\"feat3\", \"feat4\"]]
      2 ln.save(features)
----> 3 feature_set = ln.FeatureSet.from_values(features)

File ~/PycharmProjects/lamindb/lamindb/_feature_set.py:136, in from_values(cls, values, field, type, name, mute, organism, source, raise_validation_error)
    134     type = NUMBER_TYPE
    135     logger.debug(\"setting feature set to 'number'\")
--> 136 validated = registry.validate(values, field=field, mute=mute, organism=organism)
    137 values_array = np.array(values)
    138 validated_values = values_array[validated]

File ~/PycharmProjects/lamindb/lamindb/_can_curate.py:87, in validate(cls, values, field, mute, organism, source)
     75 @classmethod  # type: ignore
     76 @doc_args(CanCurate.validate.__doc__)
     77 def validate(
   (...)
     84     source: Record | None = None,
     85 ) -> np.ndarray:
     86     \"\"\"{}\"\"\"  # noqa: D415
---> 87     return _validate(
     88         cls=cls, values=values, field=field, mute=mute, organism=organism, source=source
     89     )

File ~/PycharmProjects/lamindb/lamindb/_can_curate.py:259, in _validate(cls, values, field, mute, using_key, organism, source)
    256         logger.warning(msg)
    257     return np.array([False] * len(values))
--> 259 result = validate(
    260     identifiers=values,
    261     field_values=field_values,
    262     case_sensitive=True,
    263     mute=mute,
    264     field=field,
    265 )
    266 if return_str and len(result) == 1:
    267     return result[0]

File ~/PycharmProjects/lamin-utils/lamin_utils/_inspect.py:134, in validate(identifiers, field_values, case_sensitive, mute, field, **kwargs)
    131     mute = not kwargs.get(\"logging\")
    132 import pandas as pd
--> 134 _check_type_compatibility(identifiers, field_values)
    136 identifiers = list(identifiers)
    137 identifiers_idx = pd.Index(identifiers)

File ~/PycharmProjects/lamin-utils/lamin_utils/_inspect.py:198, in _check_type_compatibility(identifiers, field_values)
    192 id_type, value_type = (
    193     _get_type_category(id_sample),
    194     _get_type_category(value_sample),
    195 )
    197 if id_type != value_type:
--> 198     raise TypeError(
    199         f\"Type mismatch: identifiers are '{id_type}' but field_values are '{value_type}'.\"
    200     )

TypeError: Type mismatch: identifiers are 'unknown' but field_values are 'str/categorical'."
}

Creating the issue here but it's across lnschema-core (fix examples) & lamin-utils

[Zethson]

Nevermind, think it does work.