Synaptic Misfirings

There's no doubt that GTA IV in its execution and scope deserves the 10's that EDGE et al have been giving it. I'm hugely impressed by the city simulation, and somewhat less impressed by the highly derivative story. So far GTA IV has not delivered one of those shiver down the spine moments that accompanied my personal seminal moments in gaming. GTA IV is for me a small increment over the state of the art. It is without a doubt a extremely well crafted experience and yet there are things that pull it back all the time. The frequently puerile humour I guess is included to appease the < 18 year old members of its fan base. I can understand why R* do this, they do it for economic reasons, apparently most GTA players don't play for the story (I guess we'll have to see if that follows for IV) and yet GTA IV has a OK story (not BioShock level writing but better than most) one that apparently wants to be taken seriously. Which is a shame because with every narrative step forward it then precedes to take two back with endless titty or dick jokes.

Hopefully one day R* or more likely Bethesda etc will really acknowledge that a substantial and increasing percentage of their audience are growing up and really want more from their virtual worlds. The 18 cert on GTA is there because the content is adult, but sadly most of what constitutes 'adult' for GTA is 'porn shop' humour rather than adult in the sense of really really exploring ideas or having anything more than cheap-shot social commentary jokes. So I'll be burning through GTA IV's story to clear the decks for "Metal Gear Solid 4: Guns of the Patriots" in June, at least that title by all accounts has some real depth to it.

I have a passion for left of field 3D packages and tools and was very happy to be introduced to TopMod in the latest issue of '3D World' magazine. This is well worth checking out, you really can make some beautiful and intricate images very easily. Much like Blender, TopMod is a classic example of how Open Source occasionally turns up real gems. I just wish the content creation software genre was better represented in the Open Source movement. Still do check out TopMod...

I'm probably horribly late to the party with this but I figured I'd share it anyway. It's one of those things that with hindsight falls into the "Doh. Obviously!" category: System.Linq.Enumerable.Aggregate use is not just limited to numeric style operations.

There are various ways of mapping a method over an array in either C# 2.0 or C# 3.0, for example:

   string[] data = { "Softimage|XSI", "Blender", "Luxology modo" };
   StringBuilder builder = new StringBuilder();
   foreach (string s in data)
      builder.Append("(").Append(s).Append(")");

   Console.WriteLine(builder.ToString());

or using the static ForEach method on Array:

   string[] data = { "Softimage|XSI", "Blender", "Luxology modo" };
   StringBuilder builder = new StringBuilder();
   Array.ForEach(data, s => builder.Append("(").Append(s).Append(")"));
   Console.WriteLine(builder.ToString());

However, nicer still if you can use C# 3.0 is System.Linq.Enumerable.Aggregate:

   string[] data = { "Softimage|XSI", "Blender", "Luxology modo" };
   Console.WriteLine(data.Aggregate(new StringBuilder(), (b, s) => b.Append("(").Append(s).Append(")")).ToString());

Not only does Aggregate being a extension method benefit from the whole "extension method pattern matching" mechanism which allows you to swap in different implementations (or specializations <smile>) depending on what names spaces you have open, but (and this is the really cool bit) it allows you to treat the whole loop operation as a value. It may not be immediately obvious if you haven't done much Linq before but the return type of Aggregate as used above is StringBuilder! This is how we can pipeline it right into a call to Console.WriteLine. Note also that Array.ForEach is limited to SysArrays whereas Linq.Enumerable.Aggregate by definition works with the more primitive IEnumerable<T> - in other words it will work just a nicely with Lists or anything else that reduces to a IEnumerable<T>. Of course the other nice advantage to using Aggregate is the lazy (pull) evaluation when walking the collection.

Aside: I've deliberately avoided talking about using the classic "for(int i = 0; i < data.Length; i++)" iteration method because I didn't want the main point to be lost behind the implementation dependent fact that the jitter can translate a classic-for into very efficient native code, safely omitting the bounds checking. This is about abstraction above the micro optimization level...

I have always maintained that Video Games have the potential to be more than just entertainment, the good ones transcend the medium (much like those sublime 16 bit AV demos back on the Atari ST and Amiga...). Video Games as a valid art form? Absolutely. They are true left brain / right brain creations. So I've always made a point of getting those games that score over 80% almost regardless of genre (OK with the exception of sports titles) because I want to experience for myself just how the developer has managed to push the medium. Now I find myself in a tricky situation with GTA IV. I don't dig the whole anti-society thing. I don't dig the car jacking and murder for money thing. I've never bought a GTA title... until now, because damn it, GTA IV is getting stunning reviews from folks who's opinion I trust. Let's be clear about this, I'm not anti violence in video games, I'm a huge first person shooter fan - but even within that genre I prefer a good degree of 'reality separation' preferring the Halo's and Unreal's to the Call of Duty's. Increasingly I get uncomfortable about being given the ability to 'shoot up' the real world - it's the same in movies for me, I just don't go to 'gun p0rn' movies like Rambo etc (although it's probably unfair to compare GTA IV to the utter crap that by all accounts the latest Rambo movie was). By all accounts GTA IV does have an actual plot and does have very good characterization. It'll be interesting to see if the game does differentiate between players who want to create the minimum of havoc vs. those who just want to drive over every pedestrian in sight (because really, if the later is your idea of a Good Time you'll probably be just as well served by Carmageddon... now there's a game for Rambo fans...)

Over the past week I made a real effort to really read 'cover to cover' (so to speak) Bjarne Stroustrup's "Evolving a language in and for the real world: C++ 1991-2006" paper. I'd skimmed it before but never really read it in the way it deserves until now. I'm so glad I did - because it's contributed greatly to my decision to spend a > 0% of home coding time on C++ rather than just exclusively coding C# 3.0 stuff at home. It's very hard to criticise anything in such a paper given the kudos of the author - that said, I do wish he would refrain from being quite so derogatory towards Microsoft and C# / .NET in particular. Then again, it's his paper so I guess he's allowed the odd back stab now and then - and to be honest the occasions when he does flash the blade left me smiling rather than angry.

Stroustrup says something in this paper which is an echo of something I've always held close to my heart and encourage other developers to aspire to, namely developers should Own the call stack. That's not to mean that you have to comprehend the full details of the problem domain at each level of the call stack but you should be able to deal with the syntax and semantics the solution-implementation is presented in. For .NET developers this means you need to know in addition to your favourite managed language some C++ and some native assembler. He goes on to point out just where C++ fits in the tool chain of languages such as Java and C#, namely "My compiler compiled your runtime" - you ignore that at your peril.

Putting aside owning the call stack for a moment, the real thing that caught me in this paper was the multi paradigm stuff. I've written before here about my increasing concern that not everything is a object and it's often a mistake to try and express things in OO terms, well Stroustrup points out again and again that OO is just one of the programming paradigms that C++ supports, he goes further to say that OO is not the reason for C++'s being, nor is it the most important programming paradigm the language supports. I'm predicting that this year I'll be having my code reviewed by someone who protests that my implementation is "Not OO enough!" - to which my answer will be: "Oh really? Let's talk about that..."

If like me you've spent some time in "Managed Code Happy Land" I'd totally recommend reading the paper, carefully, fully and giving yourself time to think about what Stroustrup has to say. I'm not u-turning here, since my day job is programming on Window's I'll likely still recommend C# for most things, but, and it's a big but I think I might be using C++/CLI a lot more than I used to. I used to joke to a mate of mine that I'd never go to a job that required me to spend >= 50% of my time doing C++, this paper (amongst other things I've recently read) means I'd probably want to revisit that statement to something >= 75% of my time.

There's a great, but all together too short, interview with Bjarne Stroustrup in the April 2008 issue of MSDN Magazine. Lots of good stuff in there - including a book recommendation I plan to follow up - but one of the things that stood out was Bjarne's comments on the use of Hungarian notation:

   "And no, don't use Hungarian. Hungarian is an awful idea. The source code should reflect the meaning of the program, not simulate a type system. If you really, really feel the need for Hungarian, you are probably using a language that is unsuitable for your application."

I'll admit that when I first started programming professionally, some 15 years ago now, I was a strong advocate and user of Hungarian. These days I limit myself to a few scope notations ("m_" for a instance members and "s_" for static members) and I don't use any type prefix notations including combinations-of or variations-on 'p' prefixes to denote pointers. I don't distinguish between native and managed languages - my position is that Hungarian has no place in either. If you're relying on the presence of Hungarian to alert developers that they've switched from .cs to .cpp files then you've got a bigger problem on your hands. It's my impression that folks who do continue to advocate heavy use of type notations do so because they pretty much live in fear of the code they work on. They neither trust the existing code, or (and this is the damning bit) the new code they add. To be honest I've seen zero indication that Hungarian contributes to a lower defect count, indeed the only thing Hungarian succeeds in lowering is the abstraction level (which is not what we want to strive to do in our code) furthermore it acts as a barrier to refactoring. Of course we're also assuming here that developer has used Hungarian correctly and consistently, which is seldom the case. I'd much rather developers put the effort into choosing decent variables names that are closer to the problem domain than slavishly writing comp.sci type prefixes. Decent variables names reduce (but obviously not eliminate) the need for commenting and make you think about the actual problem more - where as slavishly following some anachronistic type notation convention too frequently seems to accompany a school of programming typified by fear, uncertainty and doubt. Just my 2p worth... :)

First off I'm very aware that I'm about 15 years late with this insight but I still find it amazing just how many developers of the Object Orientated school are so hostile to the idea that functions still remain a equally valid and often preferable means of constructing software rather than their sacred objects. More and more I see code with classes that exist for no other reason than to act as namespaces to what amount to little more than static member functions. Fair enough languages like C# pretty much force one to express things this way, but that's a syntactic deficiency - what concerns me more is that a lot of developers seem to view functions (in the classical sense) as something lower down the programming evolutionary ladder than classes and objects. For some time I've been looking around for mainstream examples I could site to support the counter argument that objects and functions are two different and complimentary tools - and of course a prime example was right under my nose, to whit: STL. Here's a classic interview with Alex Stepanov that puts into words the first class status of functions (frequently in conjunction with and complimentary to, objects) in software construction:

   http://www.sgi.com/tech/stl/drdobbs-interview.html

Over the past 18 months or so I've personally 're-discovered' functions and have found it a liberating experience, I no longer feel the OO guilt of "I should be striving for member functions at all times" as a overriding rule when designing and writing software.

Having traded in all my Xbox 360 games along with the wretched console itself I now find myself with a whole bunch of GAME points to burn through and last Friday picked up a copy of Digital Extreme's new 'Dark Sector' game on PlayStation 3. I recall the early preview coverage in EDGE of this game before the '360 and Playstation 3 where even released, back then 'Dark Sector' was a set in space survival horror game. Some years later and it appears to have returned to Earth, Eastern Europe to be precise, having mutated into a chimeric 'Gears of War' meets 'Resident Evil' mash up. To cut to the chase, it is immense fun to play. Yeah, it's pretty shallow in terms of plot, somewhat formulaic granted, but it does deliver on the entertainment front. I have to mention the games graphics engine too which for once isn't the Unreal 3 engine but a custom in-house engine developed by Digital Extreme's themselves, said engine produces some of the nicest visuals I've seen on the PlayStation 3 to date. I honestly think in visual terms it deserves to be ranked up there alongside Drake's Fortune and Ratchet and Crank: Tools of Destruction. Whilst playing it last night I was thinking "I remember when visuals like this where reserved for the cut scenes and here I am running and gunning interactively inside them... wow." There's also a Softimage XSI connection too by way of Pendulum Studio's work on the cinematics for the game.

Over the past week I've been working my way through Digital-Tutors excellent 'Introduction to XSI 6' training and can totally recommend it to anyone who wants to rapidly get up to speed with Softimage's XSI Foundation 6. The title is project based and by the end you'll pretty much be able to reproduce the cover image - and in so doing will have been gently introduced to a good deal of XSI at a basic level. The content is first class, the movies are crystal clear and the presenter is very good at keeping you motivated. 3D modelling and animation are complex topics and I'm very aware that nothing good comes without a lot of practice and basic studies. So I've set myself the goal of trying to complete at least one simple 3D life study a week for the next month or so.

Being an unabashed geek, someone who unapologetically admits at parties to writing software for a living (and loving it), I find myself increasingly gravitating towards Microsoft's Channel 9 portal for great video conversations and interviews on the art and science of software development. Channel 9 whilst obviously focusing on Microsoft platforms and development tools in particular, frequently contains material that is truly platform agnostic - so much so that I'd whole heartedly recommend the site to any *nix developer. Even if you do still subscribe to the early Jurassic view of Microsoft as a Evil Baby Eating Empire, you'd have to be particularly narrow minded not to acknowledge the quality and passion of scientists, researchers and engineers they have working there. The following three video conversations are prime examples of just this, as well as happening to happily coincide with what I think are some of the most interesting fields in computer science, namely parallel computing, garbage collection and language design.

'Burton Smith: On General Purpose Super Computing and the History and Future of Parallelism' is a great conversation, of particular interest I thought was the thinking behind why the modern field of parallel computing often makes reference to the many core problem and just how this is subtly different from the multi core problem. It's a subtle semantic difference and leads to a interesting conversation about trends in hardware design.

   http://channel9.msdn.com/Showpost.aspx?postid=382639

I have lamented before on this blog the radio silence 'imposed' in Chris Brumme since he started working on The Next Big Thing, so imagine my surprise and pleasure to find him popping up on a Channel 9 interview with Patrick Dussud in connection with the fascinating topic of Garbage Collection (you where pre-warned about how geeky this was likely to be...)

   http://channel9.msdn.com/Showpost.aspx?postid=381293 

I'm really hoping that at this years Microsoft Professional Developers Conference we get a glimpse of what Chris and various others have been working on :-)

Finally Channel 9 has been covering the recent Lang.NET 2008 Symposium and of all the interviews the following is I think by far the most interesting, insightful and entertaining. Put three hugely experienced language designers together and have them openly and frankly discuss what's good and not so good about the Common Language Runtime. The real eye opener for me here was how Microsoft appears to now think that having a baked in type system at the lowest level of the CLR is not necessarily the only way of achieving a provably safe execution environment.

   http://channel9.msdn.com/Showpost.aspx?postid=380228

It's always worth taking time to listen to Erik Meijer (and if you haven't caught his introduction to functional programming then now's the time) - but the real star of the show here is Gilad Bracha - who pulls no punches in pointing out the inadequacies of the Common Language Runtime and then turns his blade on C++ with both whit and insight. His comments on the CLR have certainly got me thinking, while his observations on C++ seem to match my own (in that it has become a language overburdened with corner cases that contains at its core some nice concepts, template meta programming for example, but in truth it is a language who's applicability and appropriate problem-space is shrinking all the time.)

The above three fields are of course not unconnected and the research dollars Microsoft and others are pouring into them represents a reaction to the reality that the physical substrate of software has changed significantly as we left the warm and comforting waters of speed-doubling largely single threaded processors and headed out into the it'll-never-happen-in-my-lifetime waters of pervasive and increasing parallel hardware. There are very few areas of modern life, society and culture that software doesn't touch or support in some way and whilst it's often been said that all the interesting problems in computing was tackled in the 70's that statement really doesn't do justice to the effort and skill it takes to take a idea and turn it into a platform on which we can build and live.