------------------------------
On Spit and Polish, and Emacs Remote REPL
September 09, 2009
Here I am, several weeks into a serious project, and what's happening? I'm getting distracted. Why is it so hard to finish projects? I think I know the answer to that question, but I don't like it.
**I hate spit-and-polish work.**
I love *low-level* work, even to the point of dealing with memory management and optimizing assembly code. It's the spit and polish, the sanding of the surface, everything that makes a product *look* and *feel* right that gets me. And I know why: not only is it menial work, I never anticipated it in the first place!
Jeff Atwood recently discussed this difficulty using Stack Overflow as an example. It's easy to oversimplify software; actually, it's *really* easy to *super extremely* simplify software.
I started coding games in Scheme over a month ago. I spend about 3 days a week on it. Now, that doesn't seem like much, but I fiercely believed that there *had* to be simple games I could make. My initial target would be the iPhone. Unfortunately, I'm not very interested in 2d, but there has to be at least one simple 3d game I can make, right?
I'm well on my way to finishing that game. However, I just want to make it clear how badly I underestimated the amount of work it would take. I started out strong; I ported Gambit Scheme to the iPhone, and whipped up some helpful tools for it.
But suddenly, weeks would go by seemingly without much progress. I thought about where my time was going, and realized all that's necessary for a complete game: I have to market myself, find/contract all of my art and models, write different levels for the game, make a menu screen, keep track of the score, write game instructions, make it look pretty, etc.
Anyway, I am not frustrated. I have simply re-learned the same mantra I've learned so many other times: software is not simple.
This all hit me when I stayed up all night trying to figure out how to add fur to one of my sheep models in Blender. I spent 5 hours, and all I came up with is this!
Suddenly I realized that I am *not* a modeler and I can't afford to spend time doing this. If I'm not careful with how I'm spending my time, I'll end up 6 months down the road with nothing to show but a creepy sheep wearing a bear coat.
So, how about some real content, you say? Here are a few things I've done recently.
iPhone game
Despite the tone of the opening paragraphs, I have successfully built most of my first iPhone app. It is completely written in Scheme, of course. I'm not going to disclose everything yet, but here's an early screenshot (without spit and polish, of course...):
Remote REPL in Emacs
I wrote a previous post about using a remote REPL for debugging/interacting with your apps. Many people have asked me if I have integrated it with Emacs yet, as that's the next logical step. Indeed I have.
This is just a teaser, though. I'm not releasing my Emacs package because Marc Feeley (creator of Gambit Scheme) and I have talked and he wants me to finish his work on SLIME integration instead (also, it's rather buggy). I agree with him. Using Gambit Scheme under SLIME will give much more power, and we've already done a lot of the work. Please stay tuned for the release of it.
Until then, here's the kind of stuff you can do with it:
">(click to view video)
Montreal
Speaking of Marc Feeley, I made a trip up to Montreal a couple weeks ago and visited the guys working on Gambit Scheme. I also went to the Scheme conference in Boston and met many other guys working on Scheme. It was a very inspirational trip. There are some *really* smart guys working on Scheme, and I hope to work closely with them over the next few years.
Marc even invited me to present at MSLUG, which was a real honor! He videotaped the meeting and if I can get my hands on it maybe I'll post it here. I presented my recent work of using Gambit on the iPhone and discussed developing native apps in Gambit Scheme using Gambit's portable C compilation and the remote REPL. More importantly, we all went out for beer afterward.
Many thanks also to Christian Jaeger who let me stay at his place.
Snow Leopard and OpenCL
I've been keeping an eye on OpenCL and I now have a chance to play with it! I just upgraded to Snow Leopard, so I have a system fully capable of running OpenCL. I'm really look forward to playing around with it. If you haven't already seen it, you should read Arstechnica's great review of the new OS, specifically the sections about Grand Central Dispatch and OpenCL.
I'm thinking of testing matrix multiplications with OpenCL. It may be more optimal to manage a Scene Graph by pushing off all of its hierarchical matrix transformations off to the GPU. All of that depends on how much overhead there is transferring the data, of course. I can't wait to blog about my results!
Until then, I'm going to sit down and enjoy a nice some good coffee.
