Chapter 7. Python Data Structures

In our examples so far, we've already seen many of the built-in Python data structures in action. You've probably also covered many of them in introductory books or tutorials. In this chapter, we'll be discussing the object-oriented features of these data structures, when they should be used instead of a regular class, and when they should not be used. In particular, we'll be covering:

  • Tuples and named tuples
  • Dictionaries
  • Lists and sets
  • How and why to extend built-in objects
  • Three types of queues

Empty objects

Let's start with the most basic Python built-in, one that we've seen many times already, the one that we've extended in every class we have created: the object. Technically, we can instantiate an object without writing a subclass:

>>> o = object()
>>> o.x = 5
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'object' object has no attribute 'x'

Unfortunately, as you can see, it's not possible to set any attributes on an object that was instantiated directly. This isn't because the Python developers wanted to force us to write our own classes, or anything so sinister. They did this to save memory; a lot of memory. When Python allows an object to have arbitrary attributes, it takes a certain amount of system memory to keep track of what attributes each object has, for storing both the attribute name and its value. Even if no attributes are stored, memory is allocated for potential new attributes. Given the dozens, hundreds, or thousands of objects (every class extends object) in a typical Python program; this small amount of memory would quickly become a large amount of memory. So, Python disables arbitrary properties on object, and several other built-ins, by default.

Note

It is possible to restrict arbitrary properties on our own classes using slots. Slots are beyond the scope of this module, but you now have a search term if you are looking for more information. In normal use, there isn't much benefit to using slots, but if you're writing an object that will be duplicated thousands of times throughout the system, they can help save memory, just as they do for object.

It is, however, trivial to create an empty object class of our own; we saw it in our earliest example:

class MyObject:
    pass

And, as we've already seen, it's possible to set attributes on such classes:

>>> m = MyObject()
>>> m.x = "hello"
>>> m.x
'hello'

If we wanted to group properties together, we could store them in an empty object like this. But we are usually better off using other built-ins designed for storing data. It has been stressed throughout this module that classes and objects should only be used when you want to specify both data and behaviors. The main reason to write an empty class is to quickly block something out, knowing we'll come back later to add behavior. It is much easier to adapt behaviors to a class than it is to replace a data structure with an object and change all references to it. Therefore, it is important to decide from the outset if the data is just data, or if it is an object in disguise. Once that design decision is made, the rest of the design naturally falls into place.

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset