CSV records can be converted to a plain sequence of arrays:
* The resulting value will be a map with a {{Code|records}} and an optional {{Code|names}} key.
* The column names will be available if {{Code|header}} option is set to {{Code|true}}.
The resulting representation will consume consumes less memory than XML-based formats, and values can be directly accessed without conversion. Thus, it is recommendable for very large inputs and for efficient ad-hoc processing.