A QA-guy blog

As we've learned in previous posts of the series - Layered (also known as Tiered) architecture is a de-facto standard of the industry. There're two major variations of this architectural pattern that one should be aware of: 3-layered architecture (described here ) and 2-layered architecture . This post is going to concentrate on 2-layered architecture . 2-layered architecture is typically seen in UI-specific test automation frameworks. The key feature of this pattern is the absence of Business-layer, so most of the business-related logic are actually to be put into Page Objects. In order to avoid " The God class " anti-pattern and still be able to contain business-related logic in Page Object, one usually uses libraries or frameworks that allow simplifying DOM-elements location, thus letting Page Objects to concentrate less on locating elements and more on business-logic itself. The good thing in this architectural pattern is that it actually allows you...

Yet another time I think I see people using word Agile wrong. Yet another guy claimed "we need to have more predictability and stability so I think we would go for Agile methodologies, Scrum in particular". Nonsense. I don't think that "predictability, stability, whatevershitability" has anything to do with Agile. If you start Agile transformation with such a goal you've already failed. Agile is about maximizing value delivery rate in a state of uncertainty, that's pretty much what it is. If your business is uncertain as hell why do you think that "predictability" with engineering teams is going to fix things? It is not. Worse, if you make engineering teams predictable while having unpredictable environment, you will miss the whole point and make things worse while still thinking you're doing fine. Nonsense. An agile team is about removing management layers between engineering and customer, not making management happier. ...

Let's consider an example of what Layered (Tiered) architecture may look like for a test automation framework. As a system under test, we're going to use this simple and neat "Todo list" application. First, let's do a brief analysis of the application. From my brief exploration I came up with the following checklist: Area/Group Things to check Comments Todos CRUD 1 Create item Basic scenario, UI-level case 1.1 Create an item with the same name The case is not valid. There's no validation for that - nothing to check really. 1.2 Create an item with an empty/not valid/malicious name While it would be a nice thing to test manually, from test automation standpoint I would expect this to be filtered on service or repository level, so either integration or unit test . 2 Mark item complete Basic scenario, UI-level case 2.1 Un-mark item complete Basic scenario, UI-level case 3 Delete item Basic scenario, UI-level case 3.1 Delete item ...

Choosing between writing one UI tests and 100 Unit test I would always prefer the second option. Why? Well, I am actually lame with UI testing, but that not the main reason. The only thing I worry about is speed. How fast we can change software? Quality - I don't care about it much cause quality standards are contextual and can change. The only thing that is certain - there're going to be changes and we most likely don't know what those changes are about. So there's only one thing that matters - how fast we can adapt to this change. Internal quality is something that actually makes us able to change fast. UI tests do not contribute to internal quality. UI tests may even make things worse. That is the reason. And I invite you to Testcon Moscow to talk about this more - I am going to present some more thoughts on this topic there.

I am getting tired of posts about how "automated testing" going to replace "manual testing". Let me offer you a simple analogy. I have a car that has lots of useful self-check lights, like "check engine", "low gas", "low battery", etc. Automated tests are similar to those lights - if a light is on, most likely something is wrong and I need a mechanic to look at my car. Does having those self-check lights let me not to visit a mechanic for human-driven check yearly? They don't. Fact that self-check light is not on does not mean my car is OK, does it? Areas covered lots with self-check may require less time to check (as risk they would be broken is lower). However, there are lots of areas that is not feasible to cover with self-checks. There're some car parts I am not even aware of, while mechanic knows they weak points and can find an issue in a couple of minutes. It is possible to have only automated checks and not have ...

"Modern software is of a lower quality that it was in a past". Maybe. Perceived quality of software may have decreased, but I don't think that "more testing" is a proper solution. More testing does not mean more quality More testing may find more issues, but not necessarily. Simply spending more time on the activity does not mean results would be better. And somebody needs to fix bugs, test bug fixes. So we can't tell that more testing means more quality but certainly means bigger costs. Software quality != No bugs Bugs matter, but that's hardly the only factor to measure quality level. There are a whole bunch of other things that matter: UX, number of features, documentation, price, delivery model, cute logos... Spending more time on testing may mean spending less time on these things. Adequate quality level So it is obvious that one can't spend 100 years testing every possible case. I think that each product has some Adequate le...

So we sit in the room and discuss the transformation plan for a company which was working in a waterfall manner for nearly a decade. The final goal is to switch to iterative development with somewhat small iteration (let's say, 3 weeks). The "only" problem we had is that we have 3 weeks if manual regression, and it does not really fit the plan. We have options: Long waterfall-like sprint, where the first half of the sprint we develop features, second half - test them Series of short sprints and one "hardening"/"release" sprint dedicated to regression when we finally decide to release things.  There's a third obvious option, which is to decrease regression time, but no-one seems to have knowledge about how it can be done in the next 5 years. So this gets dropped. "We need to have potentially shippable increment each sprint, "hardening" sprints are not Agile!" - says one man in a room. Well, that may be true. Maybe ...