For the past few weeks a number of programmers and myself have been tasked to build an initial prototype for a system rewrite project, handed to us by management. The merit of such project is a matter of discussion for another day; for now it is enough to say that the team has been given a difficult challenge, but at the same time excited about the lessons we knew we will gain from such an adventure.
There’s been several takeaways already in terms of technology know-how – dockerized applications, front-end development with Vue, repositories as application vendor dependency, microservices – just several of the things we’ve never done before.
But the great takeaway so far is the joy of literally working together, inside a room away from distractions, the team working on one task at a time, focused, taking turns writing application or test code on a single machine, continuously discussing options and experimenting until a problem is solved or until it is time to take a break. We instantly become aligned at what we want to achieve, we immediately help teammates move forward, we learn from each other’s skills and mistakes, we have fun. It’s a wonder why we’ve never done much of this before. Perhaps it’s because of working in cubicles. Perhaps it’s because there’s nearly not enough available rooms for such software development practice. Perhaps it’s because we’ve never heard anything about mob programming until recently.
I’m sure it won’t be everyday since we have remote work schedules, but I imagine the team spending more days working together like this from here on.
I spent the recent weeks of January tinkering with Docker in both Windows 7 and Mac OS. I played with it a lot because I thought it’s something that’s useful for a grand new project we have at work, and I thought that integrating our legacy application code to it would help me learn about it more. And the exercise did help me understand the tool better, including some nuances with application performance and database connections. I was able to dockerize our legacy apps too! 🙂
Some notes to remember related to the exercise:
- Windows 7 Docker Toolbox and Docker for Mac has a performance issue with volume mounts through docker-compose. Legacy apps composed of a large number of files (especially with dependency directories) will run, but they will be painfully slow out-of-the-box. Fortunately for Docker for Mac users, docker-sync has an effective workaround for this problem. It involves running an in-sync container for the application code, separate from the docker-compose file. Unfortunately, I have not found any workarounds for said performance issue for Windows 7 (and perhaps Windows 8) Docker Toolbox users.
- Often we have to update the host file of our server machine so that we can run applications locally using a distinct, easy-to-remember URL through a browser. This means we need to add extra hosts to necessary docker containers too if we dockerize our apps. We can do this by using the extra_hosts command in docker-compose.
- The official Postgresql docker container does not include the pdo, pdo_pgsql, and pgsql drivers, which handles the connection between the application and the database. To install those drivers inside the official container, we’ll need to use a Dockerfile and build it from the docker-compose file with the build and context commands.
- Sometimes we have a need to copy the Postgresql DB files from a running container to set up a proper volume mount of database data from host to container. We can copy that data by using the convenient docker cp <source> <destination> command. I had this work in Docker for Mac. However, for Windows 7 Docker Toolbox users, a docker container is unable to use such copied data, perhaps because of the difference in OS between host and container, so I had to resort to restoring and backing up data every time I start and stop my application containers.
- As a tester, what Docker provides me is a convenient tool to test all sorts of interesting application configurations as much as I want to on a single machine, see if the apps break if I changed some service config, and find out which configurations work or not. I can add or remove a new service, or even update an existing service to a new version, like updating PHP from 5.6 to 7.1, and immediately see what impact it has on the apps themselves. These kinds of tests are often left to operations engineers, but I’m glad there is a now a way to do such tests on my own machine, before application changes even reach a dedicated testing server.
- Even if Docker makes it easy to setup an application development environment from scratch with docker-compose and Dockerfiles, it is still important to maintain a wiki of the necessary machine configurations a programmer needs to perform in order to reset or build the apps with only a single command, or two. Subtle things like custom docker-compose files, .env and php.ini files, host files, Nginx configs, or turning long docker commands into shortcuts with shell scripts or make commands.
- Dockerizing our legacy apps pushed me on a discussion with programmers about the ways they run said applications on their machines. Most of them actually just test code changes directly on Staging or another available development server. That speaks about one habit we have as a development team, and likely the reason why our apps are a pain to setup locally.
It was only several years ago when I started writing Selenium tests, first with Selenium IDE, then in Java with Webdriver, then in Ruby with Watir. Now I don’t write a lot of Selenium tests anymore, ever since I found out that it is often better (faster and more stable) to write automated checks for application features through the API. Or through unit tests. Selenium has its place in checking user flows or automating the UI, but only if I have to, if its value exceeds that of its costs. There lies an important lesson in automation: there is not a single tool that does it all. It is us who decides which tool to use for a particular test, and it helps to understand if a tool fits the specific use case.
And I think I’ve familiarized myself with a number of tools this year: Postman for API testing, Winium for automating Windows applications, BackstopJS for open-source visual testing, Cloud9 for cloud-based software development, Phonegap for HTML-based mobile app development, Docker for building shareable self-contained images of applications for development or testing, PHP testing tools (PHPUnit, Guzzle, Behat), and source code linting tools like Rubocop for Ruby. I’m not a master of these tools, but I know enough to be able to decide whether I need them (or not) for a particular thing I want to achieve. They’re in my tool belt.
Needless to say, I have outgrown the hype of automation. It is programming and tooling at its core. It helps us perform repeatable tasks without breaking a sweat, not limited to testing apps, if done with care. It is not easy. It can be rewarding. It all starts with a deep understanding of what definite task or problem actually needs solving.
And this year’s experiences has lead me to better grasp the nuances of software development, which is actually a problem of people, of teams and their habits and biases. Our team certainly has its defaults, some of which are not helping us get better at what we do. And as such from here on I’d like to contribute in key areas I believe our programmers have not had much time to think about because of project deadlines and resource constraints – dockerized application environments and shift-left testing – solutions for providing testing feedback earlier in the software development cycle, which in turn can help us build better apps and release faster.
So.. after about a week or so since I asked permission for read-write access to our application code repository I’m glad to say that I’m almost done with setting up a version of our apps locally on my machine. It is necessary because I first need to check my changes if they work locally before committing those changes. No code commits yet until said local apps have the same stability as our apps in Staging.
But there are no unit tests. How would I know if everything works after cloning the app repository and running the local settings? Only one option: I had to run local versions of my Staging application API tests. They’re slower than unit tests but at least they let me know if the apps work on some good enough level.
Running tests on local applications!
We’re in business! 🙂
Most of the problems that I encountered whilst setting up were database problems. That’s because no one was maintaining a small-but-updated version of the database. As told, I had to manually match which queries to run according to what problems my tests found. Not pretty, though I could say that in retrospect looking at the errors and finding the DB fix on my own was good exercise. Not elegant, but it helped me get familiarized a little bit with our applications as code.
There was also no documentation about the application and how to run them on various machines. Guides are important but README files were mostly left blank. I had to rely on programmer friends for clues about what to do next whenever I got stuck.
Such problems took time and patience to solve. I had to take notes about updating certain pieces too. Sometimes, I had to make changes to the code itself in order for some features to not fail locally. And yes, I need to remember not to accidentally commit those changes to the remote repository.
It would be nice if we can just go to some private repository and download an environment image or two which runs smoothly when integrated with the app repository. Update the code on a local machine and the environment updates automatically. Set up would have been done in a matter of minutes, not days. But, alas, that’s a problem worth solving for another day.
Last Wednesday afternoon I anxiously asked my boss for permission to make changes on our application code repository. I said I wanted to try fixing some of the reported bugs listed on our tracking system, if there are no other resources available to pass them to. I made a case about myself not posing any problems because of the code review process built into our repository management tool, that there’s no reason for me to merge any changes without getting feedback from a senior developer first.
He smiled at me and gleefully said “Go ahead. I’m not going to stop you.“, to which I beamed and heartily replied “Thanks, boss!”
This is a turning point in my software testing career, to be able to work on the application code directly as needed. It is actually one of my biggest frustrations – to not be able to find out for myself where the bug lives in the code and fix them if necessary. It’s always a pain to be able to do nothing but wait for a fix, and for a fix to be dependent on the resources available. In my head I think that I’m available and maybe I can do something, but I don’t explicitly have access to the application itself and the code that runs it so I can’t do anything until I have the rights to do so. That’s how it always been. Software testers are often not expected to fiddle with code, at least in my experience, especially in the past where automation was not yet known to be useful as a testing tool. Now that I have the skills and the permission to work on the application repository, I feel that my reach for making an impact on application quality has now expanded remarkably well.
Now bug-fixing is not software testing work in the traditional sense. But I figured there’s no harm in trying to fix bugs and learning the nitty-gritty details of how our legacy applications actually run deep in the code. I believe that learning technical stuff helps me communicate better with programmers. It helps me test applications in a more efficient manner too. Of course I have to consistently remind myself that I am a software-tester-first-programmer-second guy and have to be careful not to fill my days playing with code and forgetting to explore our applications themselves. That said, there are ideas I really want to experiment within our software development process, towards the goal of improving code quality and feedback, and I can only tinker with those ideas inside the application repository itself. Dockerized testing environments, code linting, and unit tests are three things I want to start building for our team, ideas that I consider to be very helpful in writing better code but has not been given enough priority through the years.
I think I’m still testing software, just extending the knowledge and practice of the various ways I perform testing.
I’m currently in the midst of a test code overhaul, a re-writing project of sorts. It started about a week ago and so far I’ve made considerable progress on what I’ve wanted to achieve with the rewrite, which is basically cleaner and more maintainable code, mostly in the sense of test data management and test description language. The number of tests running everyday in our Jenkins system has grown noticeably and I’ve felt that it’s been difficult to add certain tests because of how I structured the test data in the past, which I have not upgraded since then. The two possible avenues for running tests – on the UI and HTTP layers – also adds a bit of complexity and it’d be nice if I can integrate the two smoothly. It’s an interesting development because I did not plan on any re-writing to be done anytime soon but I guess at the back of my mind I knew it’ll happen eventually. And so I decided to take a step back from writing more tests and do some cleanup before it gets tougher to change things. I plan to finish everything in about a month or so.
At the moment, I’m reminded of the phases I’ve gone through in learning to code and writing automated checks in the past few years:
- Early 2014. It all begins with Selenium IDE, with giving the self some time to study the basic Selenese commands for writing automated checks and (more importantly) understand how to properly retrieve the page elements you want to manipulate.
- Mid 2014. Test management in Selenium IDE becomes difficult as the number of tests grow, hence the decision to switch to Selenium WebDriver. The only programming language background I had back then was C++, which was limited to only functions and logical/conditional operators, so I chose Java to work with to lessen the learning curve.
- Late 2014. Familiarized myself with Git, which hooked me on making daily commits and appreciating version control. Along the way I learned the concepts of classes and objects.
- All of 2015 up to Early 2016. I was in a trance, writing code daily and pushing myself to create all the automated checks that I wanted to run for our apps before every release. Tests run on the Eclipse IDE using TestNG and I was happy with what I had, except that those end-to-end tests are really slow. Running everything took overnight to finish, which was okay for my employer but annoying for me personally.
- Mid 2016. Re-writing existing tests in Ruby with Cucumber integration started off (when I found Jeff Morgan’s “Cucumber & Cheese” book online) as a side project for fun and testing my skill level in programming. And I did have buckets of fun! The experiment told me that there’s still a lot I need to practice on if I want to write better code, and it also told me that I can be more productive if I switch programming languages. There’s a bit less code to type in when writing code in Ruby than Java and I liked that, plus all the interesting libraries I can use. I switched to Sublime Text and used both Jenkins and the command-line interface more extensively too.
- Late 2016. As I was looking for ways to speed up end-to-end tests total execution, which by then takes about 4 hours to complete, I ended up exploring testing apps in the HTTP layer instead of in the UI. That took a lot of studying of how our apps actually behave under the hood, what data are being passed around, how images are actually sent, how to view pages without a browser, how redirections work, among other things. After years of testing apps via the user interface, this was such a refreshing and valuable period, and I completely wondered why I never knew such a thing existed until then. It wasn’t being taught extensively to testers, perhaps because it all depends on how the app was structured to run through an API.
And these phases brings me to now, where there’s a healthy dose of API and UI layer tests all checking major app features. It’s all good, just several pieces needing a cleanup, a little parallelization, better test description language, and great documentation. It’s all good, because the lessons in both programming and testing keep piling. The two practices differ in mindset but I think they complement each other, and I think that there’s no reason anyone can’t do both.
Last year I told myself I would take a step back from automation because I’ve already built tests with a working knowledge of Java and TestNG. Apparently, with the way things went this year, I was not satisfied after all with what I had. I went on to study Ruby with Watir and Cucumber for about half of 2016, and I learned how to run web applications without using browsers after that. And for a second year in a row I feel I overdid studying again, and maybe I’m getting too comfortable about writing test code. In any case, programming has become such a useful tool in my workflow – performing checks automatically that otherwise would have eaten up a chunk of my time and enabling me to test other things.
Some other realizations during the year:
- Re-writing test code using Ruby with Watir and Cucumber and other useful ruby gems pointed out how messy the checks I wrote last year using Java was. It helps that the Ruby language has less boilerplate syntax but my earlier mistakes are more about refactoring and design rather than about the programming language itself. I still need more deliberate practice at that.
- I’ve found out that I actually do security testing in my normal day-to-day exploratory work, but I’m not especially great at it because nobody has taught me how to perform good web security testing before. It seems that getting comfortable reading and manipulating API calls is a step towards becoming better at it.
- Teams who develop software with agility does not necessarily mean they implement scrum or kanban. If there’s good communication and teamwork, teams can actually forego many of the rituals and focus on the actual work.
- Writing automated checks with healthy test description language is an important skill.
- People only learn to solve problems when they feel that those problems are worth solving. And problems that are worth solving for me does not mean they’re worth solving for other people too, so I should never stop using perspectives.
- We can leverage REST calls with web automation to make them run a lot faster and more stable.
- Exploratory testing is a prerequisite to automation.
- Building a great team is difficult. There are a lot of factors to consider – team composition, empathy, attitude, member skill set, strengths and weaknesses, to name a few. It takes some time too.