I'm about to start coding a new project, that is a little more sophisticated than my usual tinkering. So, I've been thinking about the primary use case, sequence diagrams of the systems, the data structures and the like, and I still haven't written a line of code.<p>How do you organize yourself before starting to write code? What works and what doesn't? Why? Or, is it just better to start coding and figure it all out on the fly?
More advice: think in terms of iterations. You might be thinking of everything in terms of some holy grail platonic ideal of the project, when in fact you will hit a number of road bumps and change your mind along the way that may completely change that.<p>So start smaller. What's a key piece of functionality that might set a good foundation you can build off of? Building this has the dual benefit of giving you a cognitive "win" while starting to work the "bones" of the project into your muscle memory. You'll have a much clearer idea of what the next step you need to take is to get it to that larger end-goal.
I'm tired of working for years on hobbies that ultimately get big and I think can turn into businesses only to find I have built this huge unmaintainable edifice of untested spaghetti that I have to shelve.<p>So now, whether it's on the job or on my home project time, I get the git repo and build scripts squared away with some awesome testing framework that constantly watches the file system and rebuilds and running my test suite as I'm working on the code. This way I can do TDD and get higher quality code more consistently.
Check all the news I can find.<p>And when I'm not procrastinating, I write up a kind of spec, which I then revise into code sketches, and what I would like APIs to look like.<p>And then I fill it in. Rinse and repeat.
I think there comes a point beyond which planning can be used to avoid having the purity of your ideas corrupted by the inevitable messiness of their implementation. I've recognized it in myself, at least.<p>The only thing I can suggest is that you put your notes aside, set up a repo, branch and just write some code.
Start your project with a plan on a piece of paper and without a computer in front of you.<p>Not sure if this comment by edw519 can help you though it works for me.<p><a href="https://news.ycombinator.com/item?id=191275" rel="nofollow">https://news.ycombinator.com/item?id=191275</a>
> Why? Or, is it just better to start coding and figure it all out on the fly?<p>Yes it is. Write code towards the solution and see what it leads you. Then you might get into a "writers block" and can't decide whether data structure A or B is best. So you think about it some, decide which one you like and write more code. Then you might find out that you thought A was best but it turns out B was better. So you refactor your code base, write some more code and so on.
1. write down business requirement (Break to small stores that you can finish each story in a day)
2. write down technical requirement (over all)
3. write down expect test cases in English. (When something happen, what is expected)