Back in the 1980’s, when I used to spend way too much time playing games on my Apple IIGS (and earlier, my Apple IIe), one of my favourite games was Fortress, by SSI.
Fortress gave me a small game board where I would fight it out against one of several computer AI’s, where a game consisted of 21 turns, and whoever controlled most of the game board at the end was the winner.
One of the things I loved about Fortress was the way the AI’s got smarter with time. When you first started playing, it was easy to win, but after a few games, it became more challenging. This kept me coming back to Fortress as I felt I was playing against something that basically learnt as I did.
As a programmer/developer, my mind is rarely idle, and I always have a project on the go. In the 1994 I thought it would be neat to rewrite Fortress for the Apple IIGS, using higher resolution graphics.
As part of writing this blog post, I’ve run up my Apple IIGS environment (yes, I still have all of it) within the wonderful Sweet16 emulator and found that code:
I hadn’t realised just how much of the game I had written. I thought I’d only written a bit of the game logic, however it turns out I’d written a lot of the UI as well, as can be seen from when I ran it. The AI’s hadn’t been written but the basic building blocks were there.
The funny thing is, I have the code; I have a compiled binary that I can run, but I can’t remember how to re-compile the source code anymore. I’ve got a build script there, but my memory fails to help me out.
One of these days I should bring all that code out, and store it somewhere safer.
Around this time, I got distracted and much of my home based projects took a back seat, Fortress included. My work took me away from Apple development entirely for around 15 years.
So Fortress GS was left on a floppy disk (or two) in a box of backup floppies along with everything else.
Then, in 2012, after I’d been back developing for Apple hardware again for a few years I got the bug again, and, having recovered my entire Apple IIGS development environment from hundreds of floppies and some second hand SCSI drives (my how they’ve grown; did you notice the size of the “M2” hard drive above?), I was able revisit Fortress GS.
I ported the guts of the code to Objective-C and wrote a basic prototype to show to another developer at the time as a proof of concept. This one was really basic, but it allowed me to place moves for both sides by tapping the screen.
I showed this to a designer I knew at the time who thought the idea was great, but suggested that it would be more interesting with a hexagonal grid rather than the rectangular one.
I toyed with the idea at the time, but I did nothing with it; I had other projects happening, and I wanted to focus on my educational apps.
Moving up to 2016, and the release of the Apple TV, I launched my latest educational app, Tap Tangram (which I later launched as Classroom Math Drills), and due in part to my failure to recognise that I’d missed my target, and the complete lack of featuring by Apple, the app never gained any traction and failed at launch.
That left me wondering what to do next, and then it occurred to me to reboot the Fortress app idea once again. I’d also recently read a most-excellent blog article by @redblobgames about manipulating hex grids in software, so my mind was abuzz with what I could do with it.
Enter World of Hex, my latest, and final attempt to reimagine the classic Fortress for iOS and the Apple TV.
I started out just playing with the hexagonal grids code that I wrote as a port of the code provided by @redblobgames and getting the basic board working with the underlying move computations.
Once I’d done that, I sat down and brainstormed how I wanted the app to work; how the game would play and during this process, I asked myself:
“What if, rather than a simple rectangular grid of cells, we had a map of the world as a map of hexes?”
And then I got going.
“What if, the terrain was somehow represented in this 2D map of hexes. Rather than try to represent the 3rd dimension as a true 3rd dimension, colour the hexes to represent the terrain.”
“Hmm. how many cells?”
“Earths land surface area: 150,000,000 km2”
“If we say each hex has a real world “size” of 1km, then we need to be able to map out 150 million hexes eventually. Even if they aren’t all being used by players, we need a way to know where on the earth a hex maps to land.”
“So, what is probably easier, is to map the entire planet with hexes, and then mark some as usable land, and others as ocean, unusable land, etc. that means a lot more hexes in the database though. It means millions of hexes to cover the planet completely. too many.”
“Will performance be an issue? yes.”
And so it went; with performance an issue and no real idea at that point of how to make it all happen I went hunting for others that had build a world of hexes. I needed to get an idea of:
- Could I get the basic mechanism to work on an iPhone
- How many hex tiles would I need to build a reasonable approximation of the Earths land areas?
- How would it perform if I built a model with all those tiles?
After some searching with Google, I happened upon the wonderful Hexasphere.js by Rob Scanlon. This gave me hope. If this could be done in a browser, then I could do it.
This is where I started to hit upon the boundaries of my knowledge of 3D modelling and SceneKit. I also found myself struggling with some of the math concepts involved, having to trust in these people that obviously handle it better than I.
I did get Hexasphere working, though it was extremely slow because every hexagonal tile was being implemented as a separate SceneKit node. It did work, but it just wasn’t going to cut it for a production quality game. At this point I was using very large hexagonal tiles, so the tie count was still quite low. Once I increased the resolution of the model, there would be a lot more.
I ended up posting a question or two on the Apple developer forums and the Games Stack Exchange. These helped me better understand how to improve the performance of my 3D model however I was still hitting problems in that the on-screen representation of the Hexasphere was not high enough quality.
I spent several weeks working on it and getting some great help from colleagues who knew math, and 3D rendering far better than I. The end result of that was a perfectly rendered Hexasphere using only 4 SceneKit nodes that rendered at a full 60fps on devices as old as the iPad2. The change was to put all of those tiles into a single model, and to colour them individually via the shader and it’s inputs.
I finally had what I needed to get on with the game.
At this point it was just a matter of bringing all of the pieces of the puzzle together and making them work well.
For this game, the main pieces were:
- The hexasphere code
- The Hex Grid code
- SceneKit and SpriteKit
- CloudKit (iCloud based database)
I’ve already spent enough time on the hexasphere and hex grid, so I’ll try to restrict the rest of this post to the hurdles I had finishing off the app and bringing it all together.
SceneKit and SpriteKit
Apple’s engineers have done a wonderful job of these two API’s. Having developed most of my apps with Cocos2D, the transition to SpriteKit and SceneKit was pretty painless. The primary difference for me was the coordinate system.
The main reasons I went with Apple’s frameworks this time were:
- I wanted to be able to render the 3D world, which Cocos2D wouldn’t do.
- I also wanted to branch out and learn something new.
That said, the trick was that I needed to be able to overlay my 2D game components on top of the 3D components. After a little research I discovered that Apple had kindly given us an “easy” way to do this via the overlaySKScene property of the SCNView class.
This works remarkably well however it does introduce some problems because there are bugs in the Apple frameworks (at least, there are at the time I write this). I found that there are some things, like animations of the SpriteKit nodes that need to be forced to be done within the SceneKit renderer thread. It seems that Apple use a multi-threaded renderer for SceneKit/SpriteKit and some operations that you’d expect to be thread safe, aren’t.
With a lot of help from Apple Developer Technical Support, I found and fixed this problem and filed a bug report #32015449 (github project) accordingly.
Another issue related directly to the use of overlaySKSCene was an incompatibility with the tvOS focus engine (it basically doesn’t work). I ended up having to port a focus engine I’d written for Cocos2D on tvOS and enhance it to work with World of Hex. I’ve also filed a bug report for this issue: #30628989 (github project).
Apart from this, SceneKit and SpriteKit work a treat and have made my life so much easier.
CloudKit and iCloud Integration
Once I’d decided to expand the original game beyond a single game board, and to allow people to play games in a world of game boards I needed a way to store the game boards in the cloud so that everyone sees the same thing.
When I started to develop this idea my family and I were enjoying Pokemon GO for the novelty it provided. As a user, one of the things I really didn’t like about Pokemon GO was the way it forced users to either associate our existing Google account with the app, or to create a brand new Google account just for the game. There were other options, but they all involved forcing the user to log into a specific account, just for the game.
So I looked at Apple’s CloudKit which is just one part of the whole iCloud service layer that Apple has been building and developing for years now. One of the beauties of CloudKit is that for every person using an iPhone or iPad that is logged into iCloud, an app integrating CloudKit will just work because there’s no explicit login required.
This is what I wanted. On the whole, the CloudKit integration was very straight forward and it does just work.
I really enjoyed the ease with which Apple have allowed us to define our database structure via the CloudKit dashboard, make changes and even migrate those changes from development to production environments painlessly.
If there is one thing that I found lacking it is that in the dashboard, there is no way to simply remove all existing data without also wiping the database model itself.
World of Hex has grown far beyond what I originally set out to write. It’s nothing like my original attempt back in 1994 on the Apple IIGS, and even my really early brainstorming of last year differs somewhat from what I’ve built.
One of the reasons I build these apps is for the challenge and to keep my active mind busy. I certainly don’t make much of an income from them (though, mind you, I wouldn’t complain), so there’s a lot of satisfaction in having an idea realised and released into the world. Yes it can be crushing when it doesn’t take off, but, as I mention in the credits scene within World of Hex (can you find it?), “Never Give Up”.
Learning some of the quirks of Apple’s frameworks has certainly been a challenge. Cocos2D has been wonderful to work with over the years, and in some ways it’s more mature and easier to work with than SpriteKit, however SpriteKit’s deep integration is hard to pass up now that I’ve learnt it.
SceneKit offers some pretty amazing functionality from my point of view. I remember, as a teenager back in the early 80’s having a book with some algorithms for 3D line art animation that blew me away at the time. Being able to draw a model in your fave modelling tool, drop it into Xcode and have it on a device screen in minutes is insanely great. For developers out there that think its tough work creating an app, you have no idea how spoilt you are.
If you’ve read through all this, then thanks for staying till the end. It grew somewhat longer than I’d planned.
Here it is, my World of Hex. I hope you take the time to have a game, and that you enjoy it.