TODO

Dataframe
-Class conversions
  asMatrix, asCell, asStruct (or maybe use dataframe2matrix, ...)
-Index by row label
-Add functions for modifying colnames
-Consider keeping rownames in native format to accomodate time series similarly to xts.
 We can always "factorize" string lablels as needed (slower, but more general). Column
 labels are always stored in factor form.

- XXX indexing is still not perfect. This is a bad example:
x=dataframe({{1:4},{2:5},{3:6}},{'a','b','c'})   no, no, no
cf. 
x=dataframe({1:4,2:5,3:6},{'a','b','c'})   OK, this is what we want


Ratlab functions
-Wrapper for contingency table
-Other wrappers...

-Shared memory coupling for large data