How do people do error handling in K? Or is the data typically well-formed in real programs? Not bashing anything, I'm just genuinely curious.