[GiNaC-list] Variable Isolation

[David Narvaez]

On Thu, Dec 13, 2012 at 3:06 PM, Richard B. Kreckel <kreckel at ginac.de> wrote:
> For the linear case, check out lsolve.

Right, silly me. For the quadratic case, I ended up drafting this function

GiNaC::lst qsolve(const GiNaC::ex & quadratic_equation, const
GiNaC::symbol & solve_for)
{
	GiNaC::lst solutions;

	assert(quadratic_equation.nops() == 2);

	GiNaC::ex lhs = *quadratic_equation.begin();
	GiNaC::ex rhs = *(quadratic_equation.begin() + 1);
	GiNaC::ex solvable_ex = (lhs - rhs).expand();

	assert(solvable_ex.degree(solve_for) == 2);

	GiNaC::ex a = solvable_ex.coeff(solve_for, 2);
	GiNaC::ex b = solvable_ex.coeff(solve_for, 1);
	GiNaC::ex c = solvable_ex.coeff(solve_for, 0);

	solutions.append((-b + GiNaC::sqrt(GiNaC::pow(b, 2) - 4 * a * c)) / 2);
	solutions.append((-b - GiNaC::sqrt(GiNaC::pow(b, 2) - 4 * a * c)) / 2);

	return solutions;
}

Makes sense? How can it be improved?

David E. Narvaez