Creating an Engine from Scratch

I've been coding since I was ten. That's nearly thirty years now. In that time I've worked on games ranging from simple Apple II fare to big-budget PC and console titles. I've also written a ton of applications and tools, on the PC and Mac and that old Apple II, too.

So I can create an engine from scratch if I want to. Do I want to?

Right now, the 2D RPG I'm working on is an XNA title. I'm not using an engine, tho I am using the XNA and .NET libraries. Coding up my own engine is not too bad, but there's still tons of little cruft that I have to code. I wrote a line-wrapping routine last night; the code that breaks up a string of text across multiple lines. It's perhaps the tenth time I've written such a routine.

It's the sort of thing that makes me wish I had used an engine, something that would give me those sorts of routines. I know how to write them and I've got old code laying about that I can crib from, but I still need to modify the code for the language (C#) and platform (XNA) du jour.

The downside to using an engine is learning it. I remember old Mac documentation, back in the pre-OS X eras. It was awesome. It took a long time to read (compared to today's documentation), but it was fucking stellar. Not just explanations of parameter values, but meaningful sample code (including error-checking) for nearly every use that a particular function had.

Today, most of the documentation and comments I see are like:

int key; // the key

wtf? That comment is noise. It's a waste of space and the time it takes me to look at that and then realize it contributes NOTHING.

Engines are also often buggy; if you don't have source, you're just screwed. Wait six months for the next update and hope it fixes your problem. And being generic engines, they have a strange mix of too-much and too-little functionality. They don't do exactly what you want, but they can do a dozen things that are similar but still inappropriate for your specific needs. So it takes you days to figure out what it does, and then days trying to make it dance the way you need it to -- only to find out it doesn't do that. Then days more to figure out how to change your game spec so that your requirements can be met by the subset of functionality that the engine actually provides.

Using an engine is a great idea if it would take you too long to write it yourself, and if you can amortize the learning time across multiple projects (ie reusing the engine). Finding a great, easy-to-use, powerful, flexible, well-documented, and robust engine is a pain in the ass. Plus, generally impossible. There just aren't enough engines out there (and they're not profitable enough to develop) for them to be stellar products.

The big engines these days are cross-platform; PC, XBox, PS3, and Wii. There's some smaller engines that enjoy some niches, but I don't have high hopes for their lifetimes. Those big cross-platform engines can meet some of those important feature points, but (back when I was seriously looking at them, at game-industry day jobs) they weren't 100%. There's some smaller engines that have a lot of promise, too -- yet they're very niche.

Niche is good. If it does just what you need it to, a niche engine can be great. You don't waste time and money on features you won't use, and it's more likely that the engine does exactly what you need, does it well, and can document its small feature set well. Here, I'm thinking about engines for the iPhone, or 3D-shooter-specific engines, etc.

Which brings me back to that first point: if you lack the experience and talent to quickly develop an engine yourself, chances are, using a cumbersome, weak, inflexible, poorly-documented, and buggy engine will be your best choice. This is mostly cuz of the arcana associated with a new platform and the toolchain that goes with it. An engine isn't just line-wrapping code; it's a sound system, 3D rendering and scene graphs, advanced shaders, exporters for Max and Maya, input abstractions, UI widgets, plus tons more. Maybe even a scripting language and enough of a game shell that most of what you need to do is plug in some art assets, do a bit of scripting, and ship your product. If I was developing an XBLA or a PS3 or even a AAA PC title, I'd be using an engine.

There's a time balance between learning someone else's engine and writing the code you need yourself. The smaller the requirements you have for the engine, the more it makes sense to do it yourself. That's one reason why I'm writing my own engine. Plus, I get to reuse this engine for the next title I do. Plus, I'm not stuck with broken code. Plus, the engine does exactly what I need it to do.

It's still annoying to have to write line-wrapping code. Again.