[BUG]cstrs_self_adaptive hides the problems integer dimensions from the algorithm

[SchwarzschildX]

Describe the bug
If a constrained integer problem is solved using cstrs_self_adaptive with pagmo::sga (or any other integer capable algorithm), the decision vectors consist of floating point numbers instead of integers.

Expected behavior
Using cstrs_self_adaptive on constrained integer problems creates integer decision vectors.

Environment (please complete the following information):

• OS: Win10 x64
• Installation method: source
• Version: 2.16.0

Idea
As far as I understand, cstrs_self_adaptive creates a penalized_udp of the original problem. This penalized_udp simply must relay the get_nix to the underlying problem.

vector_double::size_type penalized_udp::get_nix() const {
    return m_pop_ptr->get_problem().get_nix();
}

[bluescarni]

Thanks for the report. I am going to ping @darioizzo for this as he wrote the algorithm originally.

[SchwarzschildX]

to summarize:

• using cstrs_self_adaptive, the original problem is encapsulated within a penalized_udp problem
• the algorithm tries to get the problems integer dimensions via a call to the problems get_nix() method
• penalized_udp has no get_nix() method implemented

[SchwarzschildX]

#466 pull request