Computed structural data of a crosstab. It basically contains the full dataset of the column axis, which then allows us to inject artificial rows into the dataset.
We have the assumption, that the data is already pre-sorted in some way and that all rows are given in that order. As the order can be arbitrary, we do not attempt to sort or assume that items are comparable. This model is only guaranteed to work well, if the data set is properly normalized. A sane MDX datasource is guaranteed to return such a normalized dataset.
@author Thomas Morgner