Basing an automated testing effort on using only a capture tool such as IBM Rational® Robot to record and play back test cases has its drawbacks. Running complex and powerful tests is time consuming and expensive when using only a capture tool. Because these tests are created ad hoc, their functionality can be difficult to track and reproduce, and they can be costly to maintain.
A better choice for an automated testing team that's just getting started might be to use a test automation framework, defined as a set of assumptions, concepts, and practices that constitute a work platform or support for automated testing. In this article I'll attempt to shed a little light on a handful of the test automation frameworks I'm familiar with -- specifically, test script modularity, test library architecture, keyword-driven/table-driven testing, data-driven testing, and hybrid test automation. I won't evaluate which framework is better or worse but will just offer a description and a demonstration of each and, where appropriate, some tips on how to implement it in the IBM Rational toolset.
The test script modularity framework requires the creation of small, independent scripts that represent modules, sections, and functions of the application-under-test. These small scripts are then used in a hierarchical fashion to construct larger tests, realizing a particular test case.
Of all the frameworks I'll review, this one should be the simplest to grasp and master. It's a well-known programming strategy to build an abstraction layer in front of a component to hide the component from the rest of the application. This insulates the application from modifications in the component and provides modularity in the application design. The test script modularity framework applies this principle of abstraction or encapsulation in order to improve the maintainability and scalability of automated test suites.
To demonstrate the use of this framework, I'll automate a simple test case for the Windows Calculator program (see Figure 1) to test the basic functions (add, subtract, divide, and multiply).
Figure 1. The Windows Calculator
At the bottom level of the script hierarchy are the individual scripts testing addition, subtraction, multiplication, and division. As examples, the first script that follows tests addition and the second script tests subtraction.
Sub Main |
Sub Main |
The two scripts at the next level of the hierarchy would then be used to represent the Standard view and the Scientific view available from the View menu. As the following script for the Standard view illustrates, these scripts would contain calls to the scripts we built previously.
'Test Script Modularity Framework |
And finally, the topmost script in the hierarchy would be the test case to test the different views of the application.
'Test Script Modularity Framework |
From this very simple example you can see how this framework yields a high degree of modularization and adds to the overall maintainability of the test suite. If a control gets moved on the Calculator, all you need to change is the bottom-level script that calls that control, not all the test cases that test that control.
The test library architecture framework is very similar to the test script modularity framework and offers the same advantages, but it divides the application-under-test into procedures and functions instead of scripts. This framework requires the creation of library files (SQABasic libraries, APIs, DLLs, and such) that represent modules, sections, and functions of the application-under-test. These library files are then called directly from the test case script.
To demonstrate the use of this framework, I'll automate the same test case as above but use an SQABasic library. The library contains a function to perform the operations. Following are the header file (.sbh) and the library source file (.sbl).
'Header File |
'Library Source File |
Using this library, the following test case script can be made.
'Test Library Architecture Framework |
From this example, you can see that this framework also yields a high degree of modularization and adds to the overall maintainability of the test suite. Just as in test script modularity, if a control gets moved on the Calculator, all you need to change is the library file, and all test cases that call that control are updated.
Keyword-driven testing and table-driven testing are interchangeable terms that refer to an application-independent automation framework. This framework requires the development of data tables and keywords, independent of the test automation tool used to execute them and the test script code that "drives" the application-under-test and the data. Keyword-driven tests look very similar to manual test cases. In a keyword-driven test, the functionality of the application-under-test is documented in a table as well as in step-by-step instructions for each test.
If we were to map out the actions we perform with the mouse when we test our Windows Calculator functions by hand, we could create the following table. The "Window" column contains the name of the application window where we're performing the mouse action (in this case, they all happen to be in the Calculator window). The "Control" column names the type of control the mouse is clicking. The "Action" column lists the action taken with the mouse (or by the tester). And the "Arguments" column names a specific control (1, 2, 3, 5, +, -, and so on).
| Window | Control | Action | Arguments |
|---|---|---|---|
| Calculator | Menu | View, Standard | |
| Calculator | Pushbutton | Click | 1 |
| Calculator | Pushbutton | Click | + |
| Calculator | Pushbutton | Click | 3 |
| Calculator | Pushbutton | Click | = |
| Calculator | Verify Result | 4 | |
| Calculator | Clear | ||
| Calculator | Pushbutton | Click | 6 |
| Calculator | Pushbutton | Click | - |
| Calculator | Pushbutton | Click | 3 |
| Calculator | Pushbutton | Click | = |
| Calculator |  | Verify Result | 3 |
This table represents one complete test; more can be made as needed in order to represent a series of tests. Once you've created your data table(s), you simply write a program or a set of scripts that reads in each step, executes the step based on the keyword contained the Action field, performs error checking, and logs any relevant information. This program or set of scripts would look similar to the pseudocode below:
Main Script / Program |
From this example you can see that this framework requires very little code to generate many test cases. The data tables are used to generate the individual test cases while the same code is reused. The IBM Rational toolset can be extended using interactive file reads, queries, or datapools, or you can use other tools (freeware, other development tools, and such) along with IBM Rational tools in order to build this type of framework.
Data-driven testing is a framework where test input and output values are read from data files (datapools, ODBC sources, cvs files, Excel files, DAO objects, ADO objects, and such) and are loaded into variables in captured or manually coded scripts. In this framework, variables are used for both input values and output verification values. Navigation through the program, reading of the data files, and logging of test status and information are all coded in the test script.
This is similar to table-driven testing in that the test case is contained in the data file and not in the script; the script is just a "driver," or delivery mechanism, for the data. Unlike in table-driven testing, though, the navigation data isn't contained in the table structure. In data-driven testing, only test data is contained in the data files.
The IBM Rational toolset has native data-driven functionality when using the SQABasic language and the IBM Rational datapool structures. To demonstrate the use of this framework, we'll test the order form from the test sample application Classics A (see Figure 2).
Figure 2. Order form from the sample application Classics A
If we record data entry into this window, we get the following:
'Data Driven Framework |
We can use datapools to set up test cases that test valid and invalid credit card numbers and expiration dates. The datapool shown in Figure 3, for example, would be for a test case that would test the date field.
Figure 3. Sample datapool for a test case that would test the date field
If we modify the script to accept this data, we get the following:
'Data Driven Framework |
I had to add SQABasic commands to manipulate the datapools. I also added a
While loop to allow for the processing of each row in the datapool. I should also mention the use of the SQABasic command UCase within the If Then statement. UCase is used to make the argument (in this case, the datapool return value) all uppercase. This way the comparisons aren't case sensitive, making the code more robust.This framework tends to reduce the overall number of scripts you need in order to implement all of your test cases, and it offers the greatest flexibility when it comes to developing workarounds for bugs and performing maintenance. Much like table-driven testing, data-driven testing requires very little code to generate many test cases. This framework is very easy to implement using the IBM Rational toolset, and there's a lot of detailed documentation available with how-tos and examples.
The most commonly implemented framework is a combination of all of the above techniques, pulling from their strengths and trying to mitigate their weaknesses. This hybrid test automation framework is what most frameworks evolve into over time and multiple projects. Figure 4 gives you an idea of how you could combine the approaches of the different frameworks within the IBM Rational toolset.
