We're done! Handed in the Android App with the team and we got an 8. It was a really fun to bring all the Java and React-Native experience together, combine the design patterns and quickly make a quite complex Android app. The best part was the teamwork though. I'm really convinced that the most important part of making great software is having a good vibe with your team. I'm not the only one though who thinks this, so does Brian Kernighan. In an epic 1980's video at Bell Labs he explains the foundation's of UNIX (on which your probably reading this) and it's pipe operator. He explains that because the programs had to be so miniscule at the time, the operating was setup with strong focus on multi user support and sharing a good vibe between everyone who's working on the system. So our little Android app built on this long tradition of UNIX, not only literally as Android is a *NIX system, but also figuratively as the team really got into the (git)flow and put down features from push notifications to firebase and fragments to location services. It was a blast playing with Android, so now it's time to check out the neighbors and polish up my Swift apps.
As always it's been a while since my last post. In the meantime we've released another app and have two more in the works. I've gotten to know React-native more in-depth and it's latest version has really ironed out a lot of issues, especially thanks to the auto-linking of packages. That being said, the Java certification is also in the final stretch and has made me excited to build a fully native Swift app in addition to the Java one we are creating for the certification. I was surprised to find that Swift has far more features and complexity to it than Java. Not only do you have far more control over variable declaration, with options like weak, guard, get and set, built in. But also the protocol and delegation pattern was at first a bit steep to understand. This is also due to the fact that Swift seems to throw terms I know in a blender, a let is a const and an interface is a protocol for instance. There's always I big discussion about hybrid vs native, in terms of speed. Hybrid is faster to develop but harder to maintain, Native runs faster but is slower to develop. Why I'm interested in building fully native apps though is simply the tooling that comes with it. Both Apple and Google invest millions in creating a highly optimized eco-system for there app developers and I would like to be able to benefit fully from all these new features directly and as soon as they come out. With hybrid you will always have to wait for it to trickle down and then hope that the person who made the package keeps maintaining it. By going fully native you have the newest features as soon as the come out and you can count on Apple and Google for timely updates. Anyway that's all for now, lots to do and very excited about being done with the certification, making some Swift and Kotlin apps and getting some sunshine. 😎
It's a few months on again and I've really grown to enjoy React a lot. The addition of hooks in the latest update (16.8) of React is one of the few times a hype in the JS world turned out actually to be well founded. After watching Dan Abramov's slick presentation, where he refactored a chunky old class into on of the svelte functional components, it was clear these things were winners simply based on the amount of code they clean up. But the genius of these hooks goes much further I think because they nudge you into thinking more functionally about your components in general. After understanding how these things 'think' you kind of want to think in the same way. For instance instead of having one big state in a component with a but of setState functions attacking it from all sides, I now like to have a bunch of mini state hooks, each managing their own little bit of state. When I want to update one of these from inside the functional component I simply call their set function. If I want them to be updated from outside (say Redux) then I hook up a useEffect to a useSelector. So now you have a nicely split up circular loop. Redux updates so the useSelector changes -> React notices so the useEffect -> so my component does some logic and I set my state with the useState. It's really impressive to see the React team introduce something that seemingly goes against the grain of the core of the framework, that turns out to be exactly what it needed without breaking a thing. (Here is the video in case you missed it https://youtu.be/dpw9EHDh2bM?t=687)
And... We're back! This time I'm really relieved as the last few months were a super intensive cocktail of excitement and exhaustion. Now that it's done though I'm relieved to have passed the block on Advanced Object Oriented Programming as well as have a really cool Augemented Reality app built, tested and ready to ship to the app stores. This finally gives me some time to leave the Java for a bit and try Swift. I found this really neat series of lectures from Stanford on iTunes U and now finally I've got the time to check them out. Aside from wanting to get deeper into the native developing (statically typed and compiled), I was drawn to the big portion of the course devoted to the Model View Controller architecture. This structure really helped us develop the augmented reality app rapidly whilst keeping things clear and scalable. For the View we used React-Native, for the Model Redux and for the Controllers we made classes with static functions. Because these functions are static we make sure the Controllers only deal with logic and all updates are dispatched to the Redux store. By doing this we keep the app neat and tidy whilst keeping state neatly in one place. When a View receives an event that needs dealing with outside a simple component we can just call the Controller and continue business as usual until the Model updates and React re-renders the view. It's a very nice architecture and I'm excited to see how they use it at Stanford in the world of Swift. 🤓
Things are going really well and I’m really happy. It’s good to remember once in a while that the main (if not only reason) to code is probably because it’s just super fun and rewarding to stick at something until it works. Be it an Arduino, an app or a data pipeline, it's like doing a puzzle where you get to pick your own pieces. I’m particularly chipper however because as of last week I also started working at Rotterdam’s premiere Creative Digital Agency: IN10. It’s pretty astounding to see the perfect match between their cultural and creative projects, technology stack and my passions and ambitions. Asides from this there is the spot of good luck that they let me get cracking on a React-native app they are building for the Kinderdijk windmills. This is great as I’ve spent the last few months exclusively studying on Object Orientation structures and doing a React & Redux course on the side. Now I get to combine the studies in practice whilst tinkering with this amazing app that even incorporates augmented reality and will be enjoyed by people from Korea to Copenhagen. Amazingly this app is built for the two concurrent app platforms using their respective native codes - Java and Swift - and all from one code base written solely in JS. It’s not as easy as copy-pasting what you learned on the web, but with a bit of React or Vue experience, React-native really does let you get you apps deployed in record time. You can check out the app right now by looking for Kinderdijk in the App-store. Again thanks for reading and talk to you soon 🤓
Now that I have gotten properly stuck into Java and enjoyed making multiple code challenges for the University, I’m really happy to share what I have learned about test-driven development. It’s funny that even though Java is the strictest language I’ve ever used, our goal is now to it even stricter by also enforcing semantic logic through tests. This practice has been a complete eye-opener and I really think it has to be a part of any solid code project. Just to give a bit of background: test driven development sets out to first define and write tests for all the things you do not want to happen in your app. You write these seemingly simple tests that control for code that is perfectly valid on a syntax level a.k.a. the compiler gobbles it up no problem, but that that does insane things on semantic a.k.a. on a real-world level does strange stuff. A nice example is when I was modelling a simple bank transaction and I had built in a test to make sure that if a customer did not have funds greater than 0 after transaction, the transaction would not go through. Funnily enough I had placed the test to high up in the transaction process which had the nasty side-effect of also apply to incoming transfers. This meant that a customer with say €20 could not receive €2000 because 20 - 2000 < 0. Building in tests that check for things as simple as this constantly helps not only yourself as your application grows and you lose sight of the details you were focussed on weeks or months ago. It also offers an amazing support to developers who may come in years later and write code that could undermine the goals you set out to solve.
After getting into the groove of modeling everything as attribute-only classes in domain models, we are now looking at extracting methods from use cases and creating sequence diagrams to model the communication between objects in a system. Just to recap the impressive speed of the stuff we are learning: in two sessions we have passed from the basic concept of objects instantiated from classes, to seeing how we can model the classes in a domain models and class diagrams along with object diagrams and sequence models. What I found a particularly interesting exercise was defining proper use cases and thendeducting methods from these use cases. It's surprisingly tricky to succinctly describe the functionality of a system without becoming too specific for a use case. The exercise is particularly good though, because often you will find yourself developing in a diverse team with different skillsets. Being able to clearly define use cases, is already helping me better agree with the instructional designers on the requirements and functionality that I need to develop. As for the UML notation, it's pretty tedious and I wonder how often I will encounter it in the wild, but I think for enterprises it's definitely a good tool to track, coordinate and document development. UML can definitely help to do this comprehensively and without too much overhead. It would be great to see a UML diagram for something like the Django framework just to get a quick overview of how the application is setup. In summary, I'm really impressed with the amount of tools and techniques there are to model systems. I think they take quite a bit of legwork up front, but are definitely a good way to save a lot of headaches down the road.
In this first week of self study for the Java Certification the material has focussed on some basic tools and practices to map the world our us in the Unified Modelling Language. This is a very perspective to think about as you start getting really philosophical really quick. In Object Oriented Programming every Object is instantiated or constructed from a Class. This Class is like the essential idea of an Object. Sounds pretty vague at first but anyone who has had a 101 in Philosophy in high school will quickly recognise Plato's cave. Take for instance the chair you are sitting on right now. You know it's a chair because it has four legs and a flat bit you can sit on, maybe it even has upholstery but it doesn't have to. It Could also be a chair if it had one leg or 10, as long as you could sit on it it would be a chair. This idea of kind of pealing away the Attributes from an Object until you get to a Class is what I've been getting to grips with this week. An important key for this process (and your sanity) is to keep a keen eye on the system you are trying to model so you don't start tying to exhaustively trying to Model every single Attribute of the world around us in a system. An interesting thing I have also come across lately online is a sense of cynicism on the future of Object Oriented Programming and how it's going to be replaced by Functional Programming. Having just started the study these aren't exactly the headlines that cause you to jump up and down clapping, but after having read in to it a bit I got quite excited. Functional programming seems to complement Object Oriented in that it can offer better concurrency and multi threading when dealing with heavy data processing. This has lead to a myriad of languages springing up to offer a sound basis for these concepts. Some like Scala build on Java other like Haskell take a fully functional approach. Obviously nothing is going to replace anything, there are just more and more techniques and tools being developed to solve different challenges. As a student I'm obviously eager to learn and in the future I will surely turn my attention to FP but for now it's fine to focus on the paradigm that has lead the last 30 years of computer programming and rests at the core of pretty much every electronic device in the world today. One thing is for sure; I've still got a little ways to go until I master OOP to such a degree that I need the benefits of FP for multi-threading my heavy data processing 🤓