This PR allows the use to create QuadraticModels with dense H and A.
A few notes:
in the QPDataDense struct, M1 != M2 so that the user can provide special matrices such as Diagonal matrices,
NLPModel's functionhess_structure! and hess_coord! generate COO coefficients for a full matrix with zeros in the upper triangle in order to pass the tests in NLPModelsTest,
I did not implement specialized functions for QPDataDense when using SlackModel because it is supposed to create many zeros, so using SlackModel on this type of QPData converts it to a QPDataCOO,
I did not implement specialized functions for QPDataDense when using presolve, I suggest this is done in a separate PR (I can open an issue)
This PR allows the use to create QuadraticModels with dense
HandA. A few notes:QPDataDensestruct,M1 != M2so that the user can provide special matrices such asDiagonalmatrices,hess_structure!andhess_coord!generate COO coefficients for a full matrix with zeros in the upper triangle in order to pass the tests in NLPModelsTest,QPDataDensewhen usingSlackModelbecause it is supposed to create many zeros, so usingSlackModelon this type of QPData converts it to aQPDataCOO,QPDataDensewhen usingpresolve, I suggest this is done in a separate PR (I can open an issue)