CODEWITHSUNDEEP
c++eigen
Raphael Sampaio
Raphael Sampaio

Reputation: 143

Type error in ternary operator in Eigen

I'm doing a class in c++ to generalize two sparse matrix solvers (SparseLU and Sparse Cholesky). When I try to use the ternary operator it says that the operand types are incompatible, but if I use the If statement, the code compiles.

Error 2 error: operand types are incompatible ("const Eigen::Solve < Eigen::SimplicialLDLT < Eigen::SparseMatrix < double, 0, int > , 1, Eigen::AMDOrdering < int > > , Eigen::Matrix < double, -1, 1, 0, -1, 1 > > " and "const Eigen::Solve < Eigen::SparseLU < Eigen::SparseMatrix < double, 0, int > , Eigen::COLAMDOrdering < int > > , Eigen::Matrix < double, -1, 1, 0, -1, 1 > > ")

eigen::VectorXd solve(eigen::VectorXd &b) {
    return is_cholesky ? cholesky.solve(b) : lu.solve(b); // ERROR
}

X

eigen::VectorXd solve(eigen::VectorXd &b) {
    if (is_cholesky) {
        return cholesky.solve(b);
    }
    else {
        return lu.solve(b);
    }
}

The whole code:

#pragma once  

#ifndef SOLVER_H
#define SOLVER_H

#include <Eigen/Core>
#include <Eigen/Sparse>
#include <Eigen/SparseLU>
#include <Eigen/SparseCholesky>

#define eigen Eigen

class Solver {
private:
    bool is_cholesky;
    eigen::SimplicialLDLT<eigen::SparseMatrix<double>> cholesky;
    eigen::SparseLU<eigen::SparseMatrix<double>> lu;

public:
    Solver(bool is_choleski) {
        this->is_cholesky = is_choleski;
    }

    eigen::ComputationInfo info() {
        return is_cholesky ? cholesky.info() : lu.info();
    }

    void compute(eigen::SparseMatrix<double> &B) {
        is_cholesky ? cholesky.compute(B) : lu.compute(B);
    }

    eigen::VectorXd solve(eigen::VectorXd &b) {
        return is_cholesky ? cholesky.solve(b) : lu.solve(b); // ERROR HERE
    }
};

#endif // SOLVER_H

Upvotes: 1

Views: 629

Answers (2)

YouJiacheng
YouJiacheng

Reputation: 687

https://tmp.mosra.cz/eigen-docs/TopicPitfalls.html#TopicPitfalls_ternary_operator

Offical Documentation on http://eigen.tuxfamily.org/ is unavailable now, see also eigen/doc/Pitfalls.dox

Solution: return is_cholesky ? cholesky.solve(b).eval() : lu.solve(b);

Upvotes: 0

Caleth
Caleth

Reputation: 62719

The rules for the type of a a ? b : c expression require that it is either the type of the b expression or the type of the c expression.

Here cholesky.solve(b) has a type different to lu.solve(b) and neither has an implicit conversion to the other. That the resulting expression will need a conversion to eigen::VectorXd is ignored. Thus the "operand types are incompatible" error.

return is_cholesky ? cholesky.solve(b) : lu.solve(b); // ERROR

Here each expression directly has to have a conversion to eigen::VectorXd, which exists.

if (is_cholesky) {
    return cholesky.solve(b);
}
else {
    return lu.solve(b);
}

You can force the type of the ?: expression to eigen::VectorXd by mentioning it explicitly, e.g. 

return is_cholesky ? eigen::VectorXd{ cholesky.solve(b) } : lu.solve(b);

Upvotes: 5

Related Questions