Customer Review

Reviewed in the United States on March 28, 2009
The authors give you a top level description of why automated software testing is highly desirable, along with detailed guidelines for doing so. The tone is very realistic, making you aware of many issues associated with the topic.

For one thing, you are cautioned to avoid the blandishments of a vendor who might suggest that her product will meet all your testing needs. In the authors' experience, there is no single tool that covers all major operating systems. The book also advises you to look at open source freeware. There is a surprising amount of good stuff freely available, that you might want to check first before considering proprietary products.

The book mentions many reasons for automation. These include manual tester fatigue. But also that some things are very difficult to test in a manual fashion. Often this could be because manual testing is at the GUI level. There could be bugs deep in the code, maybe in computational blocks.

Which also leads to the point that the "testers" for making automated tests often have a different skill set from manual testers. The latter might not be programmers. The former should be, with access to the source code [white box or grey box testing]. Because this gives them knowledge about what automated tests to write, that test critical aspects.

Naively, given the book's nature, we might expect it to say automate everything in sight. But the book's credibility is enhanced by it explaining that this is simply not economically feasible. The estimate is 40-60% of tests to be automated. Table 6.2 in the book is a list of questions that can be applied to each test, to suggest whether a test is suitable for automation. Roughly, tests that will be run often are a high candidate for automation.

The book also strongly recommends extensive unit testing. This is the lowest level of testing and bugs caught here have the best payoff in terms of minimising the cost to fix. A tight software development loop; "agile" as opposed to "waterfall"-like, though the book doesn't use these terms. Plus often unit testing might not be doable at the GUI level anyway, if the units are computational routines. So punting by not having automated unit tests and expecting manual tests to later find bugs in these units is very bad. Of coure, the book also describes higher level tests like regression and functional tests. But first do the unit tests.
9 people found this helpful
0Comment Report abuse Permalink

Product Details

3.0 out of 5 stars
12 customer ratings
$45.65+ Free shipping with Amazon Prime