How to write Java function that returns values of multiple data types?

Question

For example, I want to create a function that can return any number (negative, zero, or positive).

However, based on certain exceptions, I\'d like the function to re

Answer 1

IEEE Floating Point

Most languages will handle your specific example without Exceptions or union types because IEEE floating point includes a representation for NaN. In Java, use Double.NaN:

public static double quadratic(double a, double b, double c, int polarity) {
    double x = b*b - 4*a*c;

    // When x < 0, Math.sqrt(x) retruns NaN
    if (x < 0) {
        return Double.NaN;
    }
    return (-b + Math.sqrt(x) * polarity) / (2*a);
}

That produces your exact output that you wanted:

x = 4.0
x = 3.0
x = NaN
x = NaN

Exceptions

Exceptions are The Old Java Way of solving similar problems:

public static double quadratic(double a, double b, double c, int polarity) {
    double x = b*b - 4*a*c;
    // When x < 0, Math.sqrt(x) returns NaN
    if (x < 0) {
        throw new Exception("NaN")
    }
    return (-b + Math.sqrt(x) * polarity) / (2*a);
}

Here's your client code for an Exception.

a=1; b=-7; c=12;

// x = 4.0
try {
    System.out.println("x = " + quadratic(a, b, c, 1));
} catch (Exception iae) {
    System.out.println("Oopsie: " + iae.getMessage());
}

// x = 3.0
try {
    System.out.println("x = " + quadratic(a, b, c, -1));
} catch (Exception iae) {
    System.out.println("Oopsie: " + iae.getMessage());
}

// "invalid" coefficients.
a=4; b=4; c=16;

// Oopsie: NaN
try {
    System.out.println("x = " + quadratic(a, b, c, 1));
} catch (Exception iae) {
    System.out.println("Oopsie: " + iae.getMessage());
}

// Oopsie: NaN
try {
    System.out.println("x = " + quadratic(a, b, c, -1));
} catch (Exception iae) {
    System.out.println("Oopsie: " + iae.getMessage());
}

Union Types

To truly pass or return unrelated types to or from a method, you want Union types which Java does not really support. But Paguro provides Union Types which you can use in Java like this (using Or):

public static Or quadratic(double a, double b,
                                          double c, int polarity) {
    double x = b*b - 4*a*c;

    // When x < 0, Math.sqrt(x) retruns NaN
    if (x < 0) {
        return Or.bad("NaN");
    }
    return Or.good((-b + Math.sqrt(x) * polarity) / (2*a));
}

@Test public void testQuadradic() {
    double a, b, c;
    a=1; b=-7; c=12;

    // x = Good(4.0)
    System.out.println("x = " + quadratic(a, b, c, 1));

    // x = 3.0
    System.out.println(
            (String) quadratic(a, b, c, -1)
                    .match(good -> "x = " + good,
                           bad -> "Oopsie: " + bad));

    // "invalid" coefficients.
    a=4; b=4; c=16;

    // x = Bad("NaN")
    System.out.println("x = " + quadratic(a, b, c, 1));

    // Oopsie: NaN
    System.out.println(
            (String) quadratic(a, b, c, -1)
                    .match(good -> "x = " + good,
                           bad -> "Oopsie: " + bad));
}

Conclusion

For your specific example, just use Floating Point. For a more general solution, I find union types more useful than Exceptions. You can use union types as arguments to a method that might take two different inputs which have no common interface or ancestor. They are also more friendly to Functional Programming.