dropzeros!(::Hamiltonian) about quantica.jl HOT 1 CLOSED

This is actually quite dangerous because of the existence of ParametricHamiltonian. Unless we perform a copy of h when doing ph = parametric(h, ...), a dropzeros!(h) would spoil the alignment between the pointers computed upon creation for applying parameters to ph. We do a fast consistency check that would likely catch this. Should we, however, always copy h into ph? Given the risk of internal corruption it might be wise.

Another potential worry: if we do m = similarmatrix(h), then dropzeros!(h) and then bloch!(m, h, ϕs), the result should be correct and equally fast, although m will not have zeros dropped.

from quantica.jl.