This is just a proposal.
Just noting: Current implementation removes a row if only one of the column values of a row is empty.
I favor letting the user handle the data, preferably with pandas. The filter may also introduce an extra weight to slow the process if the data is big.