5. Object Orientation
This document was written for Zope 2.
To make the best use of Zope, you will need a grasp on the concept of object orientation, which is a software development pattern used in many programming languages (C++, Java, Python and others) and computer systems that simulate “real-world” behavior. It stipulates that you should design an application in terms of objects. This chapter provides a broad overview of the fundamentals of object orientation from the perspective of a Zope developer.
In Zope, as in other object-oriented systems, your application is designed around objects, or self-contained “bundles” of data and logic. It is easiest to describe these bundles by comparing them to other programming concepts.
In a typical, non-object-oriented application, you will have two things:
Code. For example, a typical CGI-based web application may have a bit of logic in the form of a PHP script, which retrieves employee data from a database and displays tabular data to a user.
Data. For example, you may have employee data stored in a database, such as MySQL or Oracle, on which some code performs read or change operations. This data exists almost solely for the purpose of the code that operates upon it; without this code, the data holds little to no value.
In a typical object-oriented application, however, you will have one thing, and one thing only:
Objects. Simply stated, these objects are collections of code and data wrapped up together. For example, you may have an “Employee” object that represents an employee. It will contain data about the employee, such as a phone number, name, and address, much like the information that would be stored in a database. However, the object will also contain “logic,” or code, that can manipulate and display its data.
In a non-object-oriented application, your data is kept separate from your code. But in an object-oriented application, both your data and your code are stored in one or more objects, each of which represents a particular “thing”. These objects can represent just about anything. In Zope, the Control_Panel is an object, Folders that you create are objects, and even the Zope “root folder” is an object. When you use the Zope “add list” to create a new item in the Zope Management Interface, you are creating an object. People who extend Zope by creating add-ons define their own types of objects, which are then entered in to the Zope “add list” so that you can create objects based on them. An add-on author might define a “Form” object or a “Weblog” object. Basically, anything that can be defined using a noun can be modeled as a Zope object.
As a programming methodology, object orientation allows software developers to design and create programs in terms of “real-world” things, such as Folders, Control_Panels, Forms, and Employees, instead of designing programs based around more “computerish” concepts like bits, streams, and integers. Instead of teaching the computer about our problem by descending to its basic vocabulary (bits and bytes), we use an abstraction to teach the computer about the problem in terms of a vocabulary that is more natural to humans. The core purpose of object orientation is to allow developers to create, to the largest extent possible, a system based on abstractions of the natural language of a computer (bits and bytes) into the real-world objects, like Employees and Forms, that we can understand more readily and quickly.
The concept of abstraction also encourages programmers to break up a larger problem by addressing the problem as smaller, more independent “sub-problems,” which allows developers to define and address solutions in much smaller, more feasible terms. When you design an application in terms of objects, they become the pieces that eventually define the solution to all the “sub-problems” of a particular “big” problem.
An object’s data is defined by its attributes, or pieces of data that describe aspects of the object. For example, an attribute of an Employee object might be called “phone_number,” which might contain a series of characters that represent the employee’s phone number. Other attributes of an Employee object might be “first_name,” “last_name”, and “job_title,” all of which give additional, detailed information about each Employee.
It may help to think of the set of attributes belonging to an object as a sort of “mini-database” that contains information representing the “real-world thing” that the object is attempting to describe. The complete collection of attributes assigned to an object defines that object’s state. When one or more of an object’s attributes are modified, the object is said to have changed its state.
The set of actions that an object may perform is defined by its methods. Methods are code definitions attached to an object that perform actions based on the object’s attributes. For example, a method of an Employee object named “getFirstName” may return the value of the object’s “first_name” attribute, while a method of an Employee object named “setFirstName” might change the value of the object’s “first_name” attribute. The “getTitle” method of an Employee object may return a value of “Vice President” or “Janitor, depending on which Employee object is being queried.
Methods are similar to functions in procedural languages like ‘C’. The key difference between a method and a function is that a method is “bound” to, or attached to, an object: instead of operating solely on “external” data that is passed to it via arguments, it may also operate on the attributes of the object to which it is bound.
In an object-oriented system, to do any useful work, an object is required to communicate with other objects in the same system. For example, it wouldn’t be particularly useful to have a single Employee object just sitting around in “object-land” with no way to communicate with it. It would then just be as “dumb” as a regular old relational database row, just storing some data without the ability to do much else. We want the capability to ask the object to do something useful, or more precisely: we want the capability for other objects to ask our Employee object to do something useful. For instance, if we create an object named “EmployeeSummary,” which is responsible for collecting the names of all of our employees for later display, we want the EmployeeSummary object to be able to ask a set of Employee objects for their first and last names.
When one object communicates with another, it is said to send a message to another object. Messages are sent to objects by way of the object’s methods. For example, our EmployeeSummary object may send a message to our Employee object by way of “calling” its “getFirstName” method. Our Employee object would receive the message and return the value of its “first_name” attribute. Messages are sent from one object to another when a “sender” object calls a method of a “receiver” object.
When you access a URL that “points to” a Zope object, you are almost always sending that Zope object a message. When you request a response from Zope by way of invoking a Zope URL with a web browser, the Zope object publisher receives the request from your browser. It then sends a Zope object a message on your browser’s behalf by “calling a method” on the Zope object specified in the URL. The Zope object responds to the object publisher with a return value, and the object publisher returns the value to your browser.
5.5. Classes and Instances
A class defines an object’s behavior and acts as a constructor for an object. When we talk about a “kind” of object, like an “Employee” object, we actually mean “objects constructed using the Employee class” or, more likely, just “objects of the Employee class.” Most objects are members of a class.
It is typical to find many objects in a system that are essentially similar to one another, save for the values of their attributes. For instance, you may have many Employee objects in your system, each with “first_name” and “last_name” attributes. The only difference between these Employee objects is the values contained within their attributes. For example, the “first_name” of one Employee object might be “Fred” while another might be “Jim”. It is likely that each of these objects would be members of the same class.
A class is to an object as a set of blueprints is to a house: as many houses can be constructed using the same set of blueprints, many objects can be constructed using the same class. Objects that share a class typically behave identically to one other. If you visit two houses that share the same set of blueprints, you will likely notice striking similarities: the layout will be the same, the light switches will be in the same places, and the fireplace will almost certainly be in the same location. The shower curtains might be different in each house, but this is an attribute of each particular house that doesn’t change its essential similarity with the other. It is much the same with instances of a class: if you “visit” two instances of a class, you would interact with both instances in essentially the same way: by calling the same set of methods on each. The data kept in the instance (by way of its attributes) might be different, but these instances behave in exactly the same way.
The behavior of two objects constructed from the same class is similar because they both share the same methods, which are not typically defined by an object itself, but are instead defined by an object’s class. For instance, if the Employee class defines the ‘getFirstName’ method, all objects that are members of the Employee class share that method definition. The set of methods assigned to an object’s class define the behavior of that object.
The objects constructed by a class are called instances of the class, or (more often) just instances. For example, the Zope ‘index’ page is an instance of the ‘Page Template’ class. The ‘index’ page has an ‘id’ attribute of ‘index’, while another page may have an ‘id’ attribute of ‘my_page’. However, while they have different attribute values, since they are both instances of the same class, they both behave identically. All the objects that can be administered using the ZMI are instances of a class. Typically, the classes from which these objects are constructed are defined in the add-ons created by Zope developers and community members.
It is sometimes desirable for objects to share the same essential behavior, except for small deviations. For example, you may want to create a ContractedEmployee object that has all the behavior of a “normal” Employee object, except that you must keep track of a tax identification number on instances of the ContractedEmployee class that is irrelevant for “normal” instances of the Employee class.
Inheritance is the mechanism that allows you to share essential behavior between two objects, while customizing one with a slightly modified set of behaviors that differ from or extend the other.
Inheritance is specified at the class level. Since classes define behavior, if we want to change an object’s behavior, we almost always need to change its class.
If we base our new “ContractedEmployee” class on the Employee class, but add a method to it named “getTaxIdNumber” and an attribute named “tax_id_number,” the ContractedEmployee class would be said to inherit from the Employee class. In the jargon of object orientation, the ContractedEmployee class would be said to subclass from the Employee class, and the Employee class would be said to be a superclass of the ContractedEmployee class.
When a subclass inherits behavior from another class, it doesn’t need to sit idly by and accept all the method definitions of its superclass if they don’t suit its needs: if necessary, the subclass can override the method definitions of its superclass. For instance, we may want our ContractedEmployee class to return a different “title” than instances of our Employee class. In our ContractedEmployee class, we might cause the ‘getTitle’ method of the Employee class to be overridden by creating a method within ContractedEmployee with a different implementation. For example, it may always return “Contractor” instead of a job-specific title.
Inheritance is used extensively in Zope objects. For example, the Zope “Image” class inherits its behavior from the Zope “File” class, since images are really just another kind of file, and both classes share many behavior requirements. But the “Image” class adds a bit of behavior that allows it to “render itself inline” by printing its content within HTML tags, instead of causing a file download. It does this by overriding the ‘index_html’ method of the File class.
5.7. Object Lifetimes
Object instances have a specific lifetime, which is typically controlled by either a programmer or a user of the system in which the objects “live”.
Instances of web-manageable objects in Zope, such as Files, Folders, and Page Templates, span from the time the user creates them until they are deleted. You will often hear these kinds of objects described as persistent objects. These objects are stored in Zope’s object database (the ZODB).
Other Zope object instances have different lifetimes: some object instances last for a “programmer-controlled” period of time. For instance, the object that represents a web request in Zope (often called REQUEST) has a well-defined lifetime, which lasts from the moment the object publisher receives the request from a remote browser, until a response is sent back to that browser, after which it is destroyed automatically. Zope “session data” objects have another well-defined lifetime, which spans from the time a programmer creates one on behalf of the user via code, until such time that the system (on behalf of the programmer or site administrator) deems it necessary to throw away the object in order to conserve space, or to indicate an “end” to the user’s session. This is defined by default as 20 minutes of “inactivity” by the user for whom the object was created.
Zope is an object-oriented development environment. Understanding Zope fully requires a grasp of the basic concepts of object orientation, including attributes, methods, classes, and inheritance, before setting out on a “for-production” Zope development project.
For a more comprehensive treatment on the subject of object orientation, buy and read The Object Primer by Scott Ambler.