Problem: - Tables with nullable `numeric` data types are breaking the `validate-data` function. - If a row has a `NaN` in Postgres, AsyncPG reads that as Python Object `Decimal('NaN')`. - Python also thinks that `Decimal('NaN') == Decimal('NaN') -> False`. Solution - Reference: https://magicstack.github.io/asyncpg/current/usage.html#example-decoding-numeric-columns-as-floats - Basically, encode and decode numeric types as strings to make things consistent.