Topic: The test framework should be application-independ

The test framework should be application-independent.

Although applications are relatively unique, the components that comprise them, in general, are not. Thus, we should focus our automation framework to deal with the common components that make up our unique applications. By doing this, we can remove all application-specific context from our framework and reuse virtually everything we develop for every application that comes through the automated test process.

---

"We should focus our automation framework to deal with the common components that make up our unique applications."

---

Nearly all applications come with some form of menu system. They also have buttons to push, boxes to check, lists to view, and so on. In a typical automation tool script there is, generally, a very small number of component functions for each type of component. These functions work with the component objects independent of the applications that contain them.

Traditional, captured automation scripts are filled with thousands of calls to these component functions. So the tools already exist to achieve application independence. The problem is, most of these scripts construct the function calls using application-specific, hard coded values. This immediately reduces their effectiveness as application-independent constructs. Furthermore, the functions by themselves are prone to failure unless a very specific application state or synchronization exists at the time they are executed. There is little error correction or prevention built-in to these functions.

To deal with this in traditional scripts we must place additional code before and\or after the command, or a set of commands, to insure the proper application state and synchronization is maintained. We need to make sure our window has the current focus. We need to make sure the component we want to select, or press, or edit exists and is in the proper state. Only then can we perform the desired operation and separately verify the result of our actions.

For maximum robustness, we would have to code these state and synchronization tests for every component function call in our scripts. Realistically, we could never afford to do this. It would make the scripts huge, nearly unreadable, and difficult to maintain. Yet, where we forego this extra effort, we increase the possibility of script failure.

What we must do is develop a truly application-independent framework for these component functions. This will allow us to implement that extra effort just once, and execute it for every call to any component function. This framework should handle all the details of insuring we have the correct window, verifying the element of interest is in the proper state, doing something with that element, and logging the success or failure of the entire activity.

We do this by using variables, and providing application-specific data to our application-independent framework. In essence, we will provide our completed test designs as executable input into our automation framework.

Does this mean that we will never have to develop application-specific test scripts? Of course not. However, if we can limit our application-specific test scripts to some small percentage, while reusing the best features of our automation framework, we will reap the rewards project after project.

click here for more details:
http://safsdev.sourceforge.net/FRAMESDataDrivenTestAutomationFrameworks.htm