is_scalar which is an issue in
package jtinormalize now returns 1 if applied to a
numericdiv(x,y) with x
being a scalar. The result was the same as div(y,x).numerics can now be marginalized. The result is the
numeric itself.as_sparta.data.frame does no longer convert to an array
first (expensive) and is therfore much more efficient now.marg is now more efficient when y is
character(0)sparta_unity_struct can now be sliced if
drop TRUE=.slicemult and div is now faster when one of the
tables is a subtale of the other. This is especially useful in
connection with Bayesian networks and the Junction Tree Algorithm.sum, normalize, get_vals and vals now
works as expected on sparta_unity classsparta_rank and
table_sizeas_array now works for sparta_unitydrop is TRUE in
slice for tables that should reduce to one row.sparsity: determines the ratio of non-zero cells and
the size of the statespace.as_df: convert a sparse table to a
data.frame. Either with or without zero cellsget_values: The same as the existing
valsunity tables now have a rank attribute which is just the unique value that all cells have. Because of this, a unity table can now be multiplied with a scalar. And regular sparta tables can be multiplied with unities of different rank. As a consequence, unities can now be marginalized also.
New function equiv to test if two sparta objects are
identical
Fixed a bug when assigning dimnames to the result of multiplication
marg, slice, mult and
div has been fixed. These functions were prone to errors
when the number of variables exceed 10.