A method to infer the datatype of a pandas Series.

infer_column_dtype.py

"""A method for inferring the data type of Series/ndarrays."""
import pandas as pd


# the type cast methods to attempt (in order)
_TYPE_CASTS = [pd.to_datetime, pd.to_numeric, pd.to_timedelta]


def infer_column_dtype(column):
        """
        Infer the datatype of a pandas column, process only if the column dtype is object.

        Args:
            column: a column as a pandas Series or NumPy ndarray

        Returns:
            the inferred data type of the column

        """
        # the column must be an object (i.e., string) to cast
        if column.dtype == "object":
            # iterate over the type cast methods (in order)
            for type_cast in _TYPE_CASTS:
                # try to cast the column using the given method
                try:
                    inferred = type_cast(column.dropna().unique())
                    return inferred.dtype
                except:
                    pass
        # return the default column dtype
        return column.dtype


# explicitly define the outward facing API of this module
__all__ = [infer_column_dtype.__name__]

Inspired by the original solution here