Explaining Properties in OOP Python

The aim of this page📝 is to simplistically illustrate Python’s concept of properties — the way to maintain/modify an object’s attributes, different from the usual getters and setters.

Here, I focus only on getter functions, also known as accessors, which are methods, conventionally named as verbs starting with get. What Python properties do is offer a different use of functionally similar implementation. For the object user/client, there is no difference between a simple instance attribute and a getter (which may compute any additional behavior before returning the value)

Our example object is a refrigerated container containing a class invariant for maximum temperature

• First, let’s look at the code

class RefrigeratedShippingContainer(ShippingContainer):

    MAX_CELSIUS = 4.0

    def __init__(self, owner_code, contents, *, celsius):
        super().__init__(owner_code, contents)
        if celsius > RefrigeratedShippingContainer.MAX_CELSIUS:
        #^^^^^^^^^^^^^^^^^^^^^establishing a class_invariant: temperature must be below max
            raise ValueError("Temperature too hot!")
            #^^^^^^^^^^^^^^^^^^^^^break if arriving celsius argument has a value > 4.0
        self.celsius = celsius

• the initializer is establishing an important class invariant that the temperature should not be below some specified maximum
• at this moment, the creation of an instance of the container object is the only way of defining the temperature
• there are 2 major handicaps:
• => 1. the temperature can be accessed only in its initial form
• => 2. the temperature cannot be changed later

In a digression, let me note the concept of invariant is a type of condition

• the term invariant is also used in the context of iteration/loops
• it is a constant condition that must hold for the iteration to proceed or object to be instantiated
• it is a constant condition for computing to proceed
• breaking the loop invariant means breaking out of the loop
• breaking the class invariant means breaking the program and raising an exception

Traditionally, the customized access to object values is implemented via Java-style getters methods (aka accessors)

• as is common, you need to create what is commonly known as backing field/property
• for more — on the etymology and comparative approach — see Explaining Backing Property in Kotlin and Java < Stack Overflow)
• in any case, Python uses underscore in this point as an implementation detail
• in addition, the underscore signifies that it is no longer to be considered to be part of the public interface
• also known as “discouraging meddling”
• for the sake of illustration of the value of the accessor, I am returning a dictionary containing both the max and current temperature
• …in other words, you may return the value in a customized/enriched/transformed format impossible with a simple instance attribute

""" THIS IS CLASSICAL JAVA IMPLEMENTATION """

class RefrigeratedShippingContainer(ShippingContainer):

    MAX_CELSIUS = 4.0

    def __init__(self, owner_code, contents, *, celsius):
        super().__init__(owner_code, contents)
        if celsius > RefrigeratedShippingContainer.MAX_CELSIUS:
            raise ValueError("Temperature too hot!")
         self._celsius = celsius

    def get_celsius(self):
        return {"max_temperature": RefrigeratedShippingContainer.MAX_CELSIUS,
                "current_temperature": self._celsius}

>>> r = oop_11_17_test.RefrigeratedShippingContainer.create_with_items("MRP", ["fish"], celsius=3)
>>> r.get_celsius()
{'max_temperature': 4.0, 'current_temperature': 3}

• theoretically, you could make a mess by changing the value of an implementation detail

Python properties provide an alternative to Java’s accessor/mutators (getters/setters)

• setters (mutators) are not discussed in this post
• properties allow getters and setters to be exposed as seemingly regular attributes
• ..in the Java world, attributes are called fields and there are no properties
• ..in C#, there are properties similar to Python, see Properties Vs Fields In C#
• anyways, properties have elevated capabilities compared to attributes
• same as with static and class methods — in Python — decorators are the basis of the property system
• instead of getters and setters, we can do something like this

class RefrigeratedShippingContainer(ShippingContainer):

    MAX_CELSIUS = 4.0

    def __init__(self, owner_code, contents, *, celsius):
        super().__init__(owner_code, contents)
        if celsius > RefrigeratedShippingContainer.MAX_CELSIUS:
            raise ValueError("Temperature too hot!")
        self._celsius = celsius

    @property
    def celsius(self):
        return {"max_temperature": RefrigeratedShippingContainer.MAX_CELSIUS,
                "current_temperature": self._celsius}

>>> r = oop_11_17_test.RefrigeratedShippingContainer.create_with_items("MRP", ["fish"], celsius=3)
>>> r.celsius
{'max_temperature': 4.0, 'current_temperature': 3}

• see, the access of a property is identical to the access to an attribute
• … → no () is required as opposed to the JAVA implementation of a getter
• … → importantly/unusually, the name of a function within the implementation is a noun
• … yet, within the implementation, property acts as a fully-fledged function

You will get an AttributeError trying to change the value of a Python property in a way you’d change the value of an instance attribute

• which may be a useful object feature, in case you are looking for creating a read-only attribute ⟹ define a property for accessing a value, but without a setter

>>> r.celsius = 5
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: can't set attribute

can't set attribute

LINKS

• http://www.java2s.com/Tutorial/SCJP/0120__Object-Oriented/JavaBeansNamingConvention.htm
• https://leetcode.com/problems/contains-duplicate/solution/#:~:text=pair%20with%20duplicates.-,Algorithm,integers.,-Complexity%20Analysis
• On Backing Property in Kotlin and Java
• Properties Vs Fields In C#