IceBreaker – PostMortem

Intro

IceBreaker is a minimalist free-pause RTS-ish thing (probably better described as an FTL-like, though bearing little similarity) set in a Cyberspace similar to the one portrayed in William Gibson’s Neuromancer (a book which changed my adolescent life and is at least partially responsible for my getting into programming).

I didn’t get much (okay, any) journal-writing done during the weekend, though there’s a vague run-down of events in the project’s github page.

So consider this (rather large) postmortem post-hoc overcompensation. (and apologies in advance for the spam)

Blender was extremely helpful for rapidly producing the future-retro look very quickly, even the sprites were tiny renderings with wireframes

You can’t quite tell, but it’s a stripped-down RTS:

no resources or buildings (instead you have gestation periods for replication)
since you can’t build unit factories, you instead have to replicate (and be vulnerable), but if you’re standing still you will heal
there /are/ classes, but they are restricted to *strength* (hit amount) and *vitality* (health)
it’s meant to be broken down into very short levels, generally with you collecting/destroying something which is being protected.

Statistics

Four litres of coffee consumed
A whole forest of tobacco
3,617 lines of code
- That’s 60 A4 pages if printed out
- According to Wolfram Alpha that’s:
  - about 17.8 metres ( 58 ft ) tall
  - 6.6 storeys high
  - and about half the diameter of the Hindenberg
- Very sore wrists (hush, you!)
somewhere between 3 and 6 hours of sleep

What went wrong

strong underlying system
- unlike my last two LudumDare attempts, I knew what I wanted to do very quickly, I wrote about three pages of ideas and then stopped when I realised I’d already made my mind up to do the first one.
  However I didn’t flesh out the details as much as usual and so started building the basic framework while pondering, knowing I could change the details later on. This resulted in a lot of code ( ~60ft worth! ) that, whilst extremely useful was probably not necessary to get the basics of the game done.
  I remain convinced that it was doable within the alotted time period (the post compo version is only an extra 4 hours work, with the last 3 mostly being unnecesary tweaking)
not enough testing of environment
- I did more preparation than previously, but I wasted time on a few things which could have been sorted out before the compo:
  - setting up the live stream stole about 1-2 hours, admittedly I was feeling a bit braindead/overwhelmed/uninspired so this was a better utilisation of time than say, nothing. But this should have “Just Worked”
  - Final builds (I’ll get to that)
using an unfamiliar framework and language (again)
- In my first LD, I used AS3/FlashPunk which I’d picked up a couple of hours before the compo. In the second, I used Java/LibGDX and didn’t complete – whilst I had familiarity with Java I was very very new to LibGDX and as a result spent wayy too much time googling. This time was a fair bit better (Haxe is quite similar to Java/AS3) but I still had little to now experience with either it, or HaxePunk
- HaxePunk is quite nice, but unfortunately not quite “there” yet for me, I wrote a disproportionately large amount of patches to the library in order to get basic features to work normally. This stole quite a bit of time, but it was far too late in the project to change ships. I look forward to using it more though.
refactoring at the halfway point
- despite having most of the system quite well designed in my head, I had to stop and write a vast swathe of code on day 2, partially to undo the odd choices of my sleep-deprived self the night before
sleep (braindead 6+6 hours)
- I should have done it sooner, and more. I’m quite good without sleep, but I ran rampant on the code-base when I started getting exhausted. Much time was spent rectifying this spaghetti. I’m not sure how long I actually slept (somewhere between 4 and 6 hours), but I easily lost 12 hours to silly choices and then the bleary-headedness upon waking.
  an early screenshot complete with pointless UI and ugly tiles
didn’t demonstrate theme clearly enough (despite following it)
- I had basic gameplay down very early in the project this time, but the sleep-spaghetti resulted in about 10-12 hours of programming which left me (effectively) where I started
planning
- I actually planned quite well in a lot of ways, but some very fundamental (and rudimentary) aspects were overlooked initially, resulting in much confusion and wasted time
submission process panic!
- I tested my environment this time to avoid this exact thing. However I discovered (at submission time) that whilst my project ran perfectly in the Flash standalone player, it would silently fail completely in-browser. It turns out all I had to do was add “-web” to the build command, but it took me far too long to discover this!
no end-game detection or automatic level progression
- despite “shipping” with a few levels, the submission process issues resulted in my missing the 20 minutes that I needed to finalise this important factor of a “short-level based game” and the gameplay suffers for it.

What went right

strong underlying system
1. Yes, it’s a dirty trick having this in both sections. But I maintain that the approach was a good one, early efforts resulted in the tutorial system being a mere 45 minutes to implement, and most new features were added extremely quickly
2. I used JSON for most of the configuration of the game, allowing rapid prototyping of enemy AI, character attributes, menus and the tutorial system)
using Haxe and SublimeText 2
1. This was a pretty awesome combination, I look forward to being able to justify the $70 license for SublimeText2 (this was my first real experience with it, and it was wonderful). I have been using (shudder) Eclipse for a while despite my lack of appreciation for IDEs in general so it was nice to have a “real” development environment again. However I’ve gotten rather dependent on Eclipse’s easy mass-refactoring, and you can really tell (names of things changed through the course of the project and thus there are some things named Agents which are actually Actors and so forth)
the game idea
1. I think this concept is pretty sound, and I enjoyed playtesting it. Definitely building some more levels and a little more “Juice” and thrusting it in the face of anyone who walks by
music and art
1. There were a few times when my brain completely went on strike, so it was good to change gears and work in Blender or Renoise to build some of the feel, having these elements in game was also fantastic for morale.
2. The music was made in about 5-15 minutes for each of the two tracks
3. Art was quite quick too, despite a few false starts
tutorial system
1. I’m really happy with the tutorial system, which could also double as a mission introduction system. It hooks into game events and each dialog of the tutorial can have a number of events required before it appears, or disappears making it very easy to make a clear (and importantly, responsive) tutorial.