Writing extra code to avoid learning new frameworks

I am a one-man shop at the place where I work, and when I started there I had zero experience and a BS degree from a below-par school of Computer Science. On top of that, my first project at the company involved not just figuring out good design principles, it also involved learning a new language. Needless to say, my code was crappy in the beginning, and all the new features I've added since then have been hacked on top of all that crappy code. It's amazing to me that my software works as well as it does.

I have learned a TON during my employment, and I am dying to refactor my code to make it more readable so future new hires can dive in and help me with it. I also REALLY want to make it easier to add new features without having to hack stuff together. I think it would be useful to learn a framework like Prism for WPF/Silverlight, but I have a huge to-do list (since I am a one-man shop), and it looks like it will take a pretty decent amount of time just to learn how to use it.

Now I have read up a little bit on Prism to where I know the basic principles behind it. Furthermore, it wouldn't be hard to write my own code that accomplishes some of the same things that Prism is used 开发者_JS百科for. I've actually done that already to some degree and I'm making good progress on making things more modular.

My question is this: should I go on writing more infrastructure code that gives me exactly what I need and no more, or should I take the time to learn something like Prism? Or maybe it could be asked like this: Should I spend time writing my own simple custom solutions, or should I spend time trying to grasp a rich, vast framework that may possibly be more complex than is necessary? And what factors should I take into account when making the decision?

I wrote my own PHP MVC framework for a recent project with exactly what I needed. It was loads of fun, taught me a lot, and an overall good experience, and I will never, ever do it again. While an excellent secondary distraction project, it highly detracted from my productivity on the main project.

Really, a lot of it depends on just how much infrastructure you will have to develop. If it's just a tiny bit that won't take more than an hour or two, go for it. If it will take significant amounts of time, use someone else's work, move on, and get your project done.

With the background you give - mostly self-educated and no peers to discuss your current development - you should absolutely check out other libraries and tools. At the very least, get new input how code can be designed and problems can be solved. You may feel that you have achieved something - and you have, congratulations - but that's a plateau, not the peak.

"I have no time to learn something new because I have so much to do"
- that's what I read in your rationale for more code. This is a warning sign - you are moving yourself into a dangerous position. No time to learn? No time to document? No time to think of all the implications? No time to do it right? No time to train a new employee? No time to call it a day?

You won't solve this problem by learning prism, or any other library, but it's the wrong rationale.

Third, code bogs you down. Having more code to maintain makes you slower. One-man-startups can crank out hundreds, even thousands of LOC per day for days and weeks. As projects and organizations get larger, you end up with an average of a few dozen.

As a recommendation from personal experience: write bulding blocks, not frameworks. Frameworks are great when you have to make the same application with different company logos over and over. Or, as TDWTF's Alex says, the key is in the differences not the similarities.

I don't want you to stop writing code, far from it. But you are discussing a tradeoff, and from the information you've given, I would recommend to put most weight on learning new things.

If the app your writing is going to be around for a while and have to be maintained, particularly by other developers, then any time spent to learn and integrate a standard framework will be worthwhile.

It'll provide documentation for how the app is written and any developer familiar with that framework will be able to pick it up faster. It should reduce the amount of code you have to write and help you concentrate on your specific business problem and not the plumbing of writing an application.

The core issue is, how many times will you re-use the framework, saving you each time the work of re-implementing similar stuff instead? Remember, the stuff you write from scratch, if it's supposed to be any good at all, will have to be tested, validated against different environments (clients &c), and maintained -- all stuff that would come to you "for free" by using a good, actively maintained framework.

If you're going to use that framework only a couple of times, maybe the net returns are still in favor of rewriting from scratch -- but if the framework covers a field that you need in more than just a couple cases, the returns on the investment of learning to use the framework (assuming it's any good!-) vs redoing things from scratch are going to be vastly positive!

I was in a similar situation when I graduated college. I received an offer from a large company about 1.5 years into my stay at the small company. What I learned was this (may be different for you and others):

1. It was an awesome idea to work at a small company right out of school. I say this because you have to wear many different hats. For example you would write the code, test the code, deploy the code, write stored procs, etc. The end result is that you are familiar with the entire process from conception to whatever. That experience is critical I think.
2. I loved writing code. I remember the days that I would be driving home and thinking of my day spent dealing with production support issues. I was spending more time supporting customers and writing "one offs" that I wasn't mostly writing code.
3. Working for a big company is a bad idea right of college. When you're working for a big company they have a specific role for you, and you have specific boundaries. If you're a developer at a big company odds are that you are not deploying the application(s) to production, or tuning stored procs.
4. Working for a big company is a great idea after working for a small one. That's because if you work at the small company it will force you to learn about a lot more than just coding. And if you understand that you will be a better developer.
5. Working with good developers makes you better. When you work with a group of guys that are good you will get better. This is because each dev has a specific history that they bring to the group, and you all learn from one another. On the group that I work with mainly right now there is: an MSBuild expert, a Silverlight expert and an F# expert, and other good guys. So some of the guys learn MSBuild from me, and I learn from them. Just talking to guys who are good can make you better.

So if I was you, don't spend too much time there. Maybe 1 or 2 years, after that find a job somewhere that has some talented developers. You will be a much better dev in 5 years. I know that I am because of my move.

I'll play contrarian: YAGNI (You Aren't Going to Need It).

What if the framework

• Is badly designed?
• Is buggy?
• Is too slow?
• Will be different in two years, and the old version won't be supported?

Discussions of frameworks often assume that frameworks are great, where the reality is that frameworks vary just like anything else, and many frameworks are larded with stuff you are never going to need.

Here's some advice that I hope bears on your more specific questions:

• Continue to make incremental improvements. It sounds like you are being productive with this approach and that it is paying off for you.

• Learn more about the framework. Or multiple frameworks. Maybe you can try a small pilot project in, say, 1 to 4 days.

• It is an honorable strategy to learn about a framework not in order to use the framework, but in order to cherry-pick the best ideas and adopt them into your own designs.

• If you decide for the time being not to adopt a framework, this is an easy decision to revisit later. If you decide to adopt a framework, backing away from it later can be very expensive. It may be worth paying some extra costs up front to reduce the probability of making a very expensive mistake.

I think where I have been burned the most is by depending on somebody else's code base that changed out from under me. I call this the "every Perl script I ever wrote was broken a year later" problem. But I work on a lot of different small projects that tend to get bursty attention and have a very long lifetime compared to the time it took to create one. If you have one big project that you're working with daily for years, you can adapt more easily to changes in external frameworks.

Who has that tool?

This is what I ask to myself everytime I need to solve every problem. This is the main factor to evaluate the effort necessary for develop the tool.

When starting a large project, everyone define well known (at project scope) usefull statements; thinking to enlarge that abstraction layer depends on the frequency of the problem, the importance of the the problem solution, the effort to develop the solution.