Python Cerberus: validating NAN and numbers using 'anyof_schema' rule

Question

I'm trying to create a Cerberus validation schema to validate a set of data in a Pandas dataframe. One of the columns should only validate on the following data:

• numbers
• NaN (this is Numpy's floating-point representation of Not a Number, and how I'm indicating null in my data)

I read the accepted answer on this post about using anyof and I structured my schema in the same way:

import numpy as np

schema = {
  'some_field': {
    'schema': {
      'anyof': [
        {
          'type': 'float',
          'allowed': [
            np.nan
          ]
        },
        {
          'type': 'number'
        }
      ]
    }
  }
}

But when I try to validate against this schema, it's returning True for documents where the field is not a number (Strings, booleans). For example:

>>> v = Validator(schema)
>>> doc = {'some_field': 'hi'}
>>> v.validate(doc)
True
>> doc = {'some_field': False}
>>> v.validate(doc)
True

It seems like this will always return true. Does anyone have any idea why it's doing this?