Number of columns is inconsistent on line #

This is more of a feature request. Currently the csvRead command assumes that the number of columns is constant.

What would be a good use case where it is not constant?