Python @property

Python has a great concept called property, which makes the life of an object oriented programmer much simpler. Before defining and going into details of what a property in Python is, let us first build an intuition on why it would be needed in the first place.

An Example To Begin With

Let us assume that one day you decide to make a class that could store the temperature in degree Celsius. It would also implement a method to convert the temperature into degree Fahrenheit. One way of doing this is as follows.

class Celsius:
    def __init__(self, temperature = 0):
        self.temperature = temperature

    def to_fahrenheit(self):
        return (self.temperature * 1.8) + 32

We could make objects out of this class and manipulate the attribute temperature, as we wished.

>>> # create new object
>>> man = Celsius()

>>> # set temperature
>>> man.temperature = 37

>>> # get temperature
>>> man.temperature
37

>>> # get degrees Fahrenheit
>>> man.to_fahrenheit()
98.60000000000001

The extra decimal places when converting into Fahrenheit is due to the floating point arithmetic error (try 1.1 + 2.2 in the Python interpreter). Whenever we assign or retrieve any object attribute like temperature, as show above, Python searches it in the object's __dict__ dictionary.

>>> man.__dict__
{'temperature': 37}

Therefore, man.temperature internally becomes man.__dict__['temperature'].

Now, let's further assume that our class got popular among clients and they started using it in their programs. They did all kinds of assignments to the object. One faithful day, a trusted client came to us and suggested that temperatures cannot go below -273 degree Celsius (students of thermodynamics might argue that it's actually -273.15), also called the absolute zero. He further asked us to implement this value constraint. Being a company that strive for customer satisfaction, we happily heeded the suggestion and released version 1.01, an upgrade of our existing class.

Using Getters and Setters

An obvious solution to the above constraint will be to hide the attribute temperature (make it private) and define new getter and setter interfaces to manipulate it. This can be done as follows.

class Celsius:
    def __init__(self, temperature = 0):
        self.set_temperature(temperature)

    def to_fahrenheit(self):
        return (self.get_temperature() * 1.8) + 32

    # new update
    def get_temperature(self):
        return self._temperature

    def set_temperature(self, value):
        if value < -273:
            raise ValueError("Temperature below -273 is not possible")
        self._temperature = value

We can see above that new methods get_temperature() and set_temperature() were defined and furthermore, temperature was replaced with _temperature. An underscore (_) at the beginning is used to denote private variables in Python.

>>> c = Celsius(-277)
Traceback (most recent call last):
...
ValueError: Temperature below -273 is not possible

>>> c = Celsius(37)
>>> c.get_temperature()
37
>>> c.set_temperature(10)

>>> c.set_temperature(-300)
Traceback (most recent call last):
...
ValueError: Temperature below -273 is not possible

This update successfully implemented the new restriction. We are no longer allowed to set temperature below -273.

Please note that private variables don't exist in Python. There are simply norms to be followed. The language itself don't apply any restrictions.

>>> c._temperature = -300
>>> c.get_temperature()
-300

But this is not of great concern. The big problem with the above update is that, all the clients who implemented our previous class in their program have to modify their code from obj.temperature to obj.get_temperature() and all assignments like obj.temperature = val to obj.set_temperature(val). This refactoring can cause headaches to the clients with hundreds of thousands of lines of codes.

All in all, our new update was not backward compatible. This is where property comes to rescue.

The Power of Property

The pythonic way to deal with the above problem is to use property. Here is how we could have achieved it.

class Celsius:
    def __init__(self, temperature = 0):
        self.temperature = temperature

    def to_fahrenheit(self):
        return (self.temperature * 1.8) + 32

    def get_temperature(self):
        print("Getting value")
        return self._temperature

    def set_temperature(self, value):
        if value < -273:
            raise ValueError("Temperature below -273 is not possible")
        print("Setting value")
        self._temperature = value

    temperature = property(get_temperature,set_temperature)

We added a print() function inside get_temperature() and set_temperature() to clearly observe that they are being executed. The last line of the code, makes a property object temperature. Simply put, property attaches some code (get_temperature and set_temperature) to the member attribute accesses (temperature). Any code that retrieves the value of temperature will automatically call get_temperature() instead of a dictionary (__dict__) look-up. Similarly, any code that assigns a value to temperature will automatically call set_temperature(). This is one cool feature in Python. Let's see it in action.

>>> c = Celsius()
Setting value

We can see above that set_temperature() was called even when we created an object. Can you guess why? The reason is that when an object is created, __init__() method gets called. This method has the line self.temperature = temperature. This assignment automatically called set_temperature().

>>> c.temperature
Getting value
0

Similarly, any access like c.temperature automatically calls get_temperature(). This is what property does. Here are a few more examples.

>>> c.temperature = 37
Setting value

>>> c.to_fahrenheit()
Getting value
98.60000000000001

By using property, we can see that, we modified our class and implemented the value constraint without any change required to the client code. Thus our implementation was backward compatible and everybody is happy.

Finally note that, the actual temperature value is stored in the private variable _temperature. The attribute temperature is a property object which provides interface to this private variable.

Digging Deeper into Property

In Python, property() is a built-in function that creates and returns a property object. The signature of this function is

property(fget=None, fset=None, fdel=None, doc=None)

where, fget is function to get value of the attribute, fset is function to set value of the attribute, fdel is function to delete the attribute and doc is a string (like a comment). As seen from the implementation, these function arguments are optional. So, a property object can simply be created as follows.

>>> property()
<property object at 0x0000000003239B38>

A property object has three methods, getter(), setter(), and delete() to specify fget, fset and fdel at a later point. This means, the line

temperature = property(get_temperature,set_temperature)

could have been broken down as

# make empty property
temperature = property()
# assign fget
temperature = temperature.getter(get_temperature)
# assign fset
temperature = temperature.setter(set_temperature)

These two pieces of codes are equivalent.

Programmers familiar with decorators in Python can recognize that the above construct can be implemented as decorators. We can further go on and not define names get_temperature and set_temperature as they are unnecessary and pollute the class namespace. For this, we reuse the name temperature while defining our getter and setter functions. This is how it can be done.

class Celsius:
    def __init__(self, temperature = 0):
        self._temperature = temperature

    def to_fahrenheit(self):
        return (self.temperature * 1.8) + 32

    @property
    def temperature(self):
        print("Getting value")
        return self._temperature

    @temperature.setter
    def temperature(self, value):
        if value < -273:
            raise ValueError("Temperature below -273 is not possible")
        print("Setting value")
        self._temperature = value

The above implementation is both, simple and recommended way to make properties. You will most likely encounter these types of constructs when looking for property in Python.

Well that's it for today.