multi table ds -> single table ds

Has anyone written any code to convert a multi table dataset into a single
table ds? So basically de-normalizing the dataset. Crystal (in its
infinite wisdom) has extremely poor performance when using a multi table
dataset for its data source, but is much better with a single de-normalized
table in a ds. I need to test the performance using a generic class that
doesn't know how many tables are in the original dataset, as well as, not
knowing what the joins are, and reproduce another dataset with a single
table.
Any help will be greatly appreciated.
gv