The importance of value and flow are the heart of day to day running at Codeweavers. During my time I've read books such as The Goal and The Toyota Way in which the likes of the Theory of Constraints and the Toyota Production System are discussed. I was also lucky enough to visit Toyota to see these practices in play.
- Pair Programming
A big part of our day to day work is done via pair programming. There are huge benefits to pair programming, but being a graduate within an organisation where pair programming is the norm is a huge benefit. Graduates usually enter the workplace with no experience in the business domain and limited technical experience. Being set loose initially would be a disaster. Graduates therefore typically shadow other developers often spending anywhere from weeks to months until they able to commit any code without supervision. With Codeweavers this was not an issue. From day one I was committing code despite being the least knowledgeable member of the team. In fact, my earliest memory of my first day involved fixing a defect in a codebase of which I knew nothing about. Thanks to my partner, this lack of knowledge was not an issue.
- Feature Toggles
The concept of a Minimal Marketable Feature (MMF) can be a catch 22 scenario. We want to develop tiny incremental features for rapid feedback, yet our customers want feature X in its entirety. In order to deliver features in this manner the concept of feature toggles are essential. Essentially code which will toggle a feature on or off is often deployed with any new features we provide. This means if something goes wrong we can instantly disable the feature without a new deployment being required. Likewise the ability to toggle features enables half finished features to be demoed to a customer within a live environment. Once a feature is deemed stable, these toggles can be removed.
- Work in Parallel
Somewhat tied to feature toggles is the concept of working in parallel or being "system green". In other words the mainline trunk from which we develop from should be deployable at any time. A benefit of working in this manner is if a feature needs disabling the old code is always ready to be enabled, likewise if an emergency fix needs deploying this can be delivered as soon as possible. There are day to day benefits from working in parallel as well. By making changes in parallel, at any one time the checked out code is only briefly in a state of being un-buildable, if we developed in a big bang approach, there could be hundreds upon hundreds of compiler errors to wade through. The important concept when working in parallel is to ensure the old code which is being extended, or replaced is cleared away once the new feature is live. The last thing we want is old code rotting with developers too afraid to delete it.
- Best Practices
It is fair to say that all the developers at Codeweavers want to develop the best code they can. Due to this practices such as SOLID and other best practices are discussed, carried out and encouraged. While I'd say that via books, the internet and other resources I have been exposed to these practices working in a industry scenario makes these concepts much more important and realistic. We all know what a "perfect" solution would be, however with deadlines and other limiting factors sometimes a more pragmatic solution is required, despite the inner perfectionists inside us all wanting to spend hours refactoring to a better solution.
There have been times when I have out rightly failed. One of my biggest regrets comes from a feature in which I let the database schema dictate the business logic. Needless to say as the business requirements evolved, the code which was so dependent on how the database persisted the data became near impossible to refactor without a rewrite. This rewrite never occurred due to deadlines meaning the feature had to go live despite myself feeling rather ashamed at how badly things had become. Despite this failure, it proved to be a huge learning experience and something I do not plan on repeating. Every time I have failed, it has helped.
- Starting Again - "a graceful retreat"
Prior to Codeweavers had I been stuck on a particular task I would have slogged away at it until the problem was resolved. Often this would mean fighting my way through a task into the early hours of the morning. By applying the principle of a "graceful retreat" I can very easily delete code and start fresh. Very quickly you will be back where you were before but in a much better condition, having learned from past experiences.
- Spike Solutions
A spike solution is essentially a throw away prototype developed before we begin a MMF in order to learn more about the problem at hand, or to test various solutions. The name refers to the fact that if you were to plot your velocity on a graph, after having produced a prototype you should have reduced the risk to the project, thus implement the feature for real easier and quicker. Every time we have developed a spike solution prior to an MMF developers often comments on how it helped, either by providing a learning experience or simply enabling the feature to be split into further MMF's. It is because of these benefits I am becoming to the conclusion that every new feature should have a quick spike solution prior to production work beginning, after all it is not easy to write tests for code you are not quite sure you know how to implement. Spikes enable this confusion to be cleared away upfront, allowing the production implementation to run smoothly.
For the most part I enjoyed university immensely. With regards the educational aspects I would find that after an interesting lecture a huge sense of motivation and interest within the subject. Spending additional time outside of a lecture discussing concepts with others was both valuable and fun. Having finished university I am a strong believer that I should maintain this inspiration and motivation. Thankfully Codeweavers provides the ability to further ourselves, and I've been lucky to attend several conferences this past year.
Having experienced one year of Codeweavers the old saying of "Practice practice practice" still holds true. Whether it be from books, conferences or general day to day work. No one can stop learning or furthering themselves if they wish to continue day to day software development. For this very reason I'm grateful, but this rule bears repeating and I look forward to another year of practice and improvement.
Common complaints and problems that I've both encountered and hear other developers raise when it comes to the practice of Test Driven Development are: Impossible to refactor without all the tests breakingMinor changes require hours of changes to test codeTest setup is huge, slow to write and difficult to understandThe use of test doubles (mocks, stubs and fakes is confusing)Over the next three posts I will demonstrate three easy steps that can resolve the problems above. In turn this will allow developers to gain one of the benefits that TDD promises - the ability to refactor your code mercifully in order to improve code quality.StepsStop Making Everything PublicLimit the Amount of Dependencies you Use A Unit is Not Always a Method or ClassCode quality is a tricky subject and highly subjective, however if you follow the three guidelines above you should have the ability to radically change implementation details and therefore improve code quality when needed.