Update problem withouth reset workspace

[damkholer]

I am solving a problem that is iteratively updated after each solve() step.
I am searching a way to setup the problem only first time (first setup) and update the matrices value on the sequent steps (update_data_vect and update_data_mat).
The problem is that i am using sparce matrices, and even if the total problem size is fixed (Amaxsize and Pmaxsize) the number of the nonzero values inside the P and A sparce marices could vary between sequent optimization steps. This cause some problems in update_data_mat function since nonzero new element are higher of nonzero old element (even if lower than nonzero max element n*n)
At the end I have concluded that for this case I have to cleanup() the workspace and setup() again the problem at each step.
Is it correct?
Other ideas?

[goulart-paul]

If you still want to use the data updating features of the solver, then your only option here is to define the initial A and P matrices such that every entry that could possibly be non-zero is included in the matrix sparsity pattern at initialization.

Note that it's not a problem to define a sparse matrix with a non-zero entry marked in the sparsity pattern (i.e. a structural non-zero), but for the numerical value of that entry to be zero. You are then free to assign these entries to be numerically non-zero in subsequent solves.