the optimal value and solution don't seem right about osqp-r HOT 1 CLOSED

The issue appears to be that OSQP is not properly receiving the P matrix. When I run your example code above, OSQP reports that nnz(A) + nnz(P) = 6, which is obviously not correct since A itself has 6 nonzeros. I think it has simply ignored P for some reason.

If I replace your P with
P <- Matrix(c(1, 0,0, 1e-10, 1,0, 0,0,1), 3, 3, sparse = TRUE)
and rerun, then I get nnz(P) + nnz(A) = 10, which is correct (counting only the upper triangle of P). I also get the solution [0.5 0.35 0.35], which also appears to be correct and matches the solution obtained from OSQP through our other language interfaces.

The underlying issue appears to be a failure to properly process P because it is of type ddiMatrix in your example, and OSQP only seems to understand it if it is of type dsCMatrix. For example, this seems to work:

model <- osqp(forceSymmetric(P), q, A, l, u, settings)
res <- model$Solve()

As does this:

P <- as(as(P,"symmetricMatrix"),"dsCMatrix")

The user should not have to do that though. I'm not very familiar with R, so I will leave this open for now and come back to it as time permits, unless someone else can propose a fix.

from osqp-r.