Confusing error during metadata validation

[npatki]

Environment Details

• SDV version: 1.0.1 (latest)
• Python version: Any
• Operating System: Any

Error Description

I expect that metadata validation should give me readable errors. Instead I get a TypeError if I accidentally input an sdtype that is not a string.

Steps to reproduce

from sdv.metadata.single_table import SingleTableMetadata

metadata = {
    "columns": {"a": {"sdtype": None}, "b": {"sdtype": "datetime"}},
    "METADATA_SPEC_VERSION": "SINGLE_TABLE_V1"
}

metadata = SingleTableMetadata.load_from_dict(metadata)
metadata.validate()

Expected Behavior: I expect a descriptive error message, similar to what happens for all other invalid sdtypes.

InvalidMetadataError: The following errors were found in the metadata:

Invalid sdtype : 'None' is not recognized. Please use one of the supported SDV sdtypes.

Observed Behavior:

TypeError: getattr(): attribute name must be string

Stack Trace

Currently observed stack trace.

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
[<ipython-input-6-f10112f21d52>](https://localhost:8080/#) in <cell line: 1>()
----> 1 metadata.validate()

4 frames
[/usr/local/lib/python3.10/dist-packages/sdv/metadata/single_table.py](https://localhost:8080/#) in validate(self)
    462         # Validate columns
    463         for column, kwargs in self.columns.items():
--> 464             self._append_error(errors, self._validate_column, column, **kwargs)
    465 
    466         if errors:

[/usr/local/lib/python3.10/dist-packages/sdv/metadata/single_table.py](https://localhost:8080/#) in _append_error(self, errors, method, *args, **kwargs)
    440         """Inplace, append the produced error to the passed ``errors`` list."""
    441         try:
--> 442             method(*args, **kwargs)
    443         except InvalidMetadataError as e:
    444             errors.append(e)

[/usr/local/lib/python3.10/dist-packages/sdv/metadata/single_table.py](https://localhost:8080/#) in _validate_column(self, column_name, sdtype, **kwargs)
    144 
    145     def _validate_column(self, column_name, sdtype, **kwargs):
--> 146         self._validate_sdtype(sdtype)
    147         self._validate_unexpected_kwargs(column_name, sdtype, **kwargs)
    148         if sdtype == 'categorical':

[/usr/local/lib/python3.10/dist-packages/sdv/metadata/single_table.py](https://localhost:8080/#) in _validate_sdtype(self, sdtype)
    137         is_default_sdtype = sdtype in self._SDTYPE_KWARGS
    138         is_anonymized_type = sdtype in SDTYPE_ANONYMIZERS
--> 139         if not (is_default_sdtype or is_anonymized_type or is_faker_function(sdtype)):
    140             raise InvalidMetadataError(
    141                 f"Invalid sdtype : '{sdtype}' is not recognized. Please use one of the "

[/usr/local/lib/python3.10/dist-packages/sdv/metadata/anonymization.py](https://localhost:8080/#) in is_faker_function(function_name)
     57     """
     58     try:
---> 59         getattr(Faker(), function_name)
     60     except AttributeError:
     61         return False

TypeError: getattr(): attribute name must be string