The second way produces more billable output tokens. Worse, when you feed that code back into the LLM service, the extra whitespace counts as input tokens.
It isn't a 1:1 substitution. Indenting each parameter like in the second example adds additional space characters: in this case, 4 extra spaces per parameter.