Tuesday, Twitch.tv personality & livestreamer Matthew “Sevadus” Zagursky showed off Grabitty LITE to his audience. I was unaware of this, and only found out after having checked out my webserver statistics, and I noticed a pretty big spike in traffic. After further investigation, it showed that almost 10.000 levels were (probably) played by his audience. It’s a shame this server doesn’t have more capacity than it has; He has coined the term “Sev-DOS”, which means that a site is taken down by all the users clicking on his links at once.

I’ve come to the conclusion that supporting multiple languages is probably something a small indie-game like this will benefit from. I’ve also decided that adding support for this early is probably a good thing, rather than trying to shoehorn it in the code later.

So this week has been spent looking at how to solve this – and trying to implement a maintainable way.

The game will contain a lot of images, which again may contain text on them; This could be anything from images of a space-bar with the text “Space” on it, a sign in the game, or something completely different; In short – we’re talking about assets. One thing I want to do, is to supply these images without the text, and have the text blended in before the texture is loaded onto the GPU. This simplifying translations significantly; also hopefully save space needed by localized versions of the game – ideally bundle them all into the same version of the game, and let the user change language as he/she wants to.

To do this, I use the FreeType-library. FreeType understands font-files, you can request a glyph of any unicode character the font provides – and you will get an 8 bit bitmap in return. This is applied on a bitmap read from a png-file before I upload it to the GPU and use it in game. This is a process I intend to only do when setting up the scene. Most guides I find on the subject talk about creating font atlas’es, which means rendering one quad for each letter you want to draw. Even though that approach will work perfectly fine, I find it unnecessary. So I’ve written some code that does this, it seems to work fine on my relatively few tests;

That’s the gist of it; Add some code for blending the font to your bitmap, hook it on to your asset loading mechanisms – and you should be good to go.

I also made a lot of work with the GUI-code, to make my job easier further down the road. The details aren’t very interesting, it’s just some simple classes that let me request different kind of primitives stacked up in a VBO – which I then later can use for drawing.

I’ve managed to cross-build the project to run on my phone – which is an Android. I plan to cross-build for IOS as well, but this makes it possible for me to weed out any GLES-related bugs early on. First thing I noticed was that the shaders failed to build because I hadn’t set precision. Happy to see that it ran fast. Not too happy about Android’s build system for NDK apps. I need to either ditch it, or make it easier to deal with – currently I’m maintaining the same set of strings in too many places.