[IndianAstronaut]

Another gotcha is variable type inference. Reading csv files can often produce varying column types. This can be a pain for any consistent data pipeline.