# Reduce algorithm complexity and promote modularity

Just about every developer has found himself in a situation where they had a complicated algorithm in a single, virtually unreadable method, that was entangled together with other methods in a class. For example, say you have a general-purpose class for solving equations:

public class EquationSolvers { public static Tuple<double, double> Quadratic(double a, double b, double c) { double disc = b*b - 4*a*c; if (disc < 0) throw new ArgumentException("Cannot solve equation with complex roots"); double sqrt = Math.Sqrt(disc); return new Tuple<double, double>( (-b + sqrt) / (2 * a), (-b - sqrt) / (2 * a)); } // other solvers here }

The above equation solver is hard-coded, meaning that to substitute a different solver, you would have to manually replace each instance. Let’s start by taking it out into a separate class. To do this, we use the Move to Another Type refactoring F6:

Then, we need to specify the class to move the method to. In order to separate concerns better, we pick a separate class called `QuadraticEquationSolver` for this:

Now that the method has been moved, let’s try taking the discriminant out to a separate calculation. This is easy — we select the discriminant calculation and invoke the Extract Method refactoring Control+Alt+M:

Now, all we need to do is to give the new method a name:

And it’s done:

private static double CalculateDiscriminant(double a, double b, double c) { return b * b - 4 * a * c; }

Now, let’s suppose that, after a while, we find a safer solver for quadratic equations. To factor it into the program, we’ll first need to create an abstract base class `QuadraticEquationSolverBase`. We use the Extract Superclass refactoring refactoring available in the Refactor This menu Control+Shift+R:

In the dialog that shows up, we get to pick which members will be promoted upwards. We only want the `CalculateDiscriminant` method:

We add an abstract definition of the `Calculate()` method (previously called `Quadratic()`) and end up with the following base class:

public abstract class QuadraticEquationSolverBase { protected double CalculateDiscriminant(double a, double b, double c) { return b*b - 4*a*c; } public abstract Tuple<double, double> Calculate(double a, double b, double c); }

We also got rid of the `static` keyword anywhere with the assumption that the implementations of `QuadraticEquationSolverBase` will be handled by a lifetime manager within our code. Consequently, ReSharper reminds us to add the `override` keyword to the renamed `Calculate` method in our QuadraticEquationSolver class:

Now, let’s say we found a safer version of the quadratic equation solver. Let’s implement it. First, we use the Create derived type context action on our base class:

Then, we are asked to implement members on this type, which we do:

Finally, we provide an implementation, making use of the base class’ `CalculateDiscriminant()` method:

class SafeQuadraticEquationSolver : QuadraticEquationSolverBase { public override Tuple<double, double> Calculate(double a, double b, double c) { double disc = CalculateDiscriminant(a, b, c); if (disc < 0) throw new ArgumentException("Cannot solve equation with complex roots"); double q = -0.5*(b + Math.Sign(b)*disc); return new Tuple<double, double> (q/a, c/q); } }

And we’re done! Now the quadratic equation solver can be easily used, with its configuration and instantiation typically handled by an IoC container.