Starting a project

It’s that time of the decade again – to start a new hobby programming project that will probably span a few years, given that it’s a side project. So I went through the usual decision tree and this time I decided to document it for my future self.

Picking a language

I really surprised myself. I picked c++. A year or so ago I would have told you I was going to use Haskell or Common Lisp for new projects. I’m pretty proficient in Python and so I wouldn’t pick that. But C++??? Now, I’ll still be doing coding challenge type tasks in common lisp this year, and next year will be the year of Haskell, but every time I touch one of these I get fingers bitten off. There is always something easy I want to do that breaks things and the solution, often from practitioners in the field, seems to be pretty messy.

Python, funnily enough, is the closest I’ve come to an ideal language for practical use. I don’t use custom defined classes very much and it’s only fake functional (it comes out in the major – no tail recursion – and in the minor – try and use lambdas with the multiprocessing module) and it can be astoundingly slow for compute intensive tasks that can’t be cast as numpy operations. But you can’t beat it in time it takes to go from a blank file to a working, self-documenting, fairly bug free program where you can often write very complex things very concisely and elegantly.

So why C++?? C++14, to be precise, drew me back from these REPL oriented languages that make prototyping so fast and easy. In a nutshell I think the standard library has turned into quite a nice thing and things like CATCH and Cmake make medium sized projects easier to manage. I anticipate an intense amount of numerical computation, potentially large amounts of generated data and an attendant desire to control the type of the data and how it’s laid out. Common Lisp, Python and, I suspect, Haskell aren’t so bothered by such low level details. I suspect I’d be able to quickly prototype what I want to do in those languages, but then I’d get into a hellhole trying to remove the bottlenecks.

Build system: Cmake: I’ve used this before and found it manageable
Command line parser: cxxopts – one file (just header)
Test system: CATCH – I’ve used this before and I like how easy it is.

One thing I’m excited to experiment with is Cling – allegedly a REPL for C++. We’ll see …

Picking a version system

This was easy. git. I toyed briefly with using this as an excuse to learn mercurial, but I decided to reduce the number of moving parts for this particular project. Mercurial will have to wait.

Picking a license

I’ve always used the GPL. This time I picked the MIT license. There were a couple of funny thoughts that went through my head as I was picking the license. Funny because they reveal very deep (but perhaps not so serious) desires that I don’t often reflect on. One dominant thought was, what if this becomes widely popular (yeah, sure) and people start to capitalize on this but I don’t get any of the upside (fame/money). You’d probably want to use the GPL to block people like that.

And so I went with the MIT license, because almost everything I know, I know because random people on the internet unconditionally shared their knowledge with me, and almost every tool I use has been unconditionally granted to me by random people on the internet.  The GPL started to sound a little mean spirited. A little too over bearing. The MIT license, on the other hand, sounded more in the hacker spirit. Here is some code. I throw it to the wind. If you find it useful send, me a post card. Just don’t sue me. Leave the lawyers out of this.

Picking a host

github, because of the network effects, and I’m using git. And to be fair, github has made a lot of effort to make the site feature full. I do hope it does not end up like sourceforge. Remember when sourceforge was where we put all our projects?

Picking a documentation system

Text files in markdown. Algorithm design notes in the code files. Because everything else gets out of date and becomes a giant interdependent mess that you deal with only if this is the day job.

Work style

Ok, this one I’m gonna get right (yeah, sure). I’m starting with the specifications, then the user manual, then the tests, then the code. This is not a magic formula, and it doesn’t stop me from skipping around, but this set of practices has always seemed more logical and a lot less irritating to me. Code is living, you always go back and change things, but as long as you follow this order when you change things, sanity prevails.

Off we go …


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s