What is a clean, Pythonic way to have multiple constructors in Python?

Python does not support explicit multiple constructors, yet there are some ways using which multiple constructors can be achieved. If multiple __init__ methods are written for the same class, then the latest one overwrites all the previous constructors and the reason for this can be, python stores all the function names in a class as key in a dictionary so, when a new function is defined with the same name, the key remains the same but the value gets overridden by the new function body.

Prerequisite – Constructors, @classmethod decorators

 Look at the example below.

Python3

class example:

    def __init__(self):
        print("One")

    def __init__(self):
        print("Two")

    def __init__(self):
        print("Three")


e = example()

Output

Three

Need for multiple constructors

Multiple constructors are required when one has to perform different actions on the instantiation of a class. This is useful when the class has to perform different actions on different parameters. The class constructors can be made to exhibit polymorphism in three ways which are listed below.

1. Overloading constructors based on arguments.
2. Calling methods from __init__.
3. Using @classmethod decorator.

This article explains how to have multiple constructors in a clean and Pythonic way with examples.

Overloading constructors based on arguments

The constructor overloading is done by checking conditions for the arguments passed and performing required actions. For example, consider passing an argument to the class sample, 

• If the parameter is an int, the square of the number should be the answer.
• If the parameter is a String, the answer should be “Hello!!”+string.
• If the parameter is of length greater than 1, the sum of arguments should be stored as the answer.

Python3

class sample:

    # constructor overloading
    # based on args
    def __init__(self, *args):

        # if args are more than 1
        # sum of args
        if len(args) > 1:
            self.ans = 0
            for i in args:
                self.ans += i

        # if arg is an integer
        # square the arg
        elif isinstance(args[0], int):
            self.ans = args[0]*args[0]

        # if arg is string
        # Print with hello
        elif isinstance(args[0], str):
            self.ans = "Hello! "+args[0]+"."


s1 = sample(1, 2, 3, 4, 5)
print("Sum of list :", s1.ans)

s2 = sample(5)
print("Square of int :", s2.ans)

s3 = sample("GeeksforGeeks")
print("String :", s3.ans)

Output

Sum of list : 15
Square of int : 25
String : Hello! GeeksforGeeks.

In the code above, the instance variable was ans, but its values differ based on the arguments. Since a variable number of arguments for the class, *args is used which is a tuple that contains the arguments passed and can be accessed using an index. In the case of int and string, only one argument is passed and thus accessed as args[0] (the only element in the tuple).

Calling methods from __init__

A class can have one constructor __init__ which can perform any action when the instance of the class is created. This constructor can be made to different functions that carry out different actions based on the arguments passed. Now consider an example : 

• If the number of arguments passed is 2, then evaluate the expression x = a²-b²
• If the number of arguments passed is 3, then evaluate the expression y = a²+b²-c.
• If more than 3 arguments have been passed, then sum up the squares, divide it by the highest value in the arguments passed.

Python3

class eval_equations:

  # single constructor to call other methods
    def __init__(self, *inp):

        # when 2 arguments are passed
        if len(inp) == 2:
            self.ans = self.eq2(inp)

        # when 3 arguments are passed
        elif len(inp) == 3:
            self.ans = self.eq1(inp)

        # when more than 3 arguments are passed
        else:
            self.ans = self.eq3(inp)

    def eq1(self, args):
        x = (args[0]*args[0])+(args[1]*args[1])-args[2]
        return x

    def eq2(self, args):
        y = (args[0]*args[0])-(args[1]*args[1])
        return y

    def eq3(self, args):
        temp = 0
        for i in range(0, len(args)):
            temp += args[i]*args[i]

        temp = temp/max(args)
        z = temp
        return z


inp1 = eval_equations(1, 2)
inp2 = eval_equations(1, 2, 3)
inp3 = eval_equations(1, 2, 3, 4, 5)

print("equation 2 :", inp1.ans)
print("equation 1 :", inp2.ans)
print("equation 3 :", inp3.ans)

Output

equation 2 : -3
equation 1 : 2
equation 3 : 11.0

 In the example above, the equation to be evaluated is written on different instance methods and made to return the answer. The constructor calls the appropriate method and acts differently for different parameters.

The expressions have been evaluated as follows:

inputs : 1,2 —> 1²-2² = 1-4 = -3

inputs : 1,2,3  —> (1² + 2²) – 3 = 5-3 = 2

inputs : 1,2,3,4,5 —> (1² + 2² + 3² + 4² + 5²) / 5 = 55/5 = 11.0

Using @classmethod decorator

This decorator allows a function to be accessible without instantiating the class. The functions can be accessed both by the instance of the class and the class itself. The first parameter of the method that is declared as classmethod is cls, which is like the self of the instance methods. Here cls refer to the class itself. This proves to be very helpful to use multiple constructors in Python and is a more Pythonic approach considered to the above ones. Consider the same example used above. Evaluate different expressions based on the number of inputs.

Python3

class eval_equations:

    # basic constructor
    def __init__(self, a):
        self.ans = a

    # expression 1
    @classmethod
    def eq1(cls, args):

      # create an object for the class to return
        x = cls((args[0]*args[0])+(args[1]*args[1])-args[2])
        return x

    # expression 2
    @classmethod
    def eq2(cls, args):
        y = cls((args[0]*args[0])-(args[1]*args[1]))
        return y

    # expression 3
    @classmethod
    def eq3(cls, args):
        temp = 0

        # square of each element
        for i in range(0, len(args)):
            temp += args[i]*args[i]

        temp = temp/max(args)
        z = cls(temp)
        return z


li = [[1, 2], [1, 2, 3], [1, 2, 3, 4, 5]]
i = 0

# loop to get input three times
while i < 3:

    inp = li[i]

    # no.of.arguments = 2
    if len(inp) == 2:
        p = eval_equations.eq2(inp)
        print("equation 2 :", p.ans)

    # no.of.arguments = 3
    elif len(inp) == 3:
        p = eval_equations.eq1(inp)
        print("equation 1 :", p.ans)

    # More than three arguments
    else:
        p = eval_equations.eq3(inp)
        print("equation 3 :", p.ans)

    #increment loop
    i += 1

Output

equation 2 : -3
equation 1 : 2
equation 3 : 11.0

In the example above, the instance of the object is not created initially. The class methods to evaluate various expression has been defined with @classmethod decorator. Now they can be called with the class name and the object is created in that class method after evaluating the expression. The instance variable holds different answers for a different number of parameters passed.