You could watch the videos, I guess. The first one(s) lays out the basic ideas, e.g it's NOT a "
Last edited:
-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
Jonathan Blow is designing his own programming language for game development
- Thread starter freakybj
- Start date
You could watch the videos, I guess. The first one(s) lays out the basic ideas, e.g it's NOT a "
I'm not defending Blow's whole position. I just didn't think the poster's specific criticism was on point. Nor does the thread need to be derailed by that discussion, does it?
He's doing a new "Sokoban
Question is, is it a more effective language or just an easier one??
If its just "easier" it could help some smaller endies or "baby 1st game" but unless is more robust than what C++ offer. I see no reason why go too deep on it then
If its just "easier" it could help some smaller endies or "baby 1st game" but unless is more robust than what C++ offer. I see no reason why go too deep on it then
One of the goals it to compile fast. He's got two backends, one in-house that outputs X86_64 binary directly, and one that goes through LLVM. So one for fast builds, one for optimized builds. As of now.
Speaking as a game programmer who has worked with Lua bindings for our engine... Lua is so much more expensive (in CPU time and memory) than C++. The footprint is massive at the scale of a AAA game.
Came looking for this point or to make it.I hope it's so intuitive that even women can learn it.
Can't support him and his back-ass thinking anymore, no matter how genius his design may be.
Good to see I wasn't the only one. I was like that's a weird ass list of choices to try and make your point with. LoL
Just watch the first video(s), and a lot of speculation will go away. Note that he himself consider them outdated, but they lay out the groundwork. If you can't bother to do that, why speculate?
I will say that initially I was very sceptical, "Oh, just what the world needs", but I've change my tune a bit. I don't agree on everything, and I worry that some of the syntax may end up sucking (he's stated it's 'temporary', but we'll see), but here's the thing.. he's taking it seriously, and he's putting in the time and the money, including hiring people to work on it.
He's not all wrong about some of the baggage we carry along in other language like C++, and the meta-programming opportunities afforded via jai looks exciting.
I will say that initially I was very sceptical, "Oh, just what the world needs", but I've change my tune a bit. I don't agree on everything, and I worry that some of the syntax may end up sucking (he's stated it's 'temporary', but we'll see), but here's the thing.. he's taking it seriously, and he's putting in the time and the money, including hiring people to work on it.
He's not all wrong about some of the baggage we carry along in other language like C++, and the meta-programming opportunities afforded via jai looks exciting.
Last edited:
Jai's Philosophy has very little in common with C#. It's actually a less feature-ful language than C++. It's supposed to be more explicit about what the machine is actually doing (less magic) and faster for getting it to do what you want, but not necessarily do more for you.
One of Thekla's upcoming games (the Sokoban-like) is written in Jai.
The problem with Rust is that its compilation constraints are not applicable to game development. Rust has a fantastic compiler, it reduces a lot of bugs that can be caused in a language like C due enforcing constraints at compile time, this is great for mission critical systems, but game development is not that. Also it has great compiler error messages, the Cargo system is great, Rust is a great language, but I don't like it for game development and I don't think its' suitable for that workflow.
Game development is a very different software engineering approach than others, architectures, techniques, data structures, etc are unique, and general purpose systems languages offer the accessibility to hardware, but are not particularly suitable to the reality of game development, such as Rust.
Rust's compiler enforcement of constraints is too resistant to the workflow of game development, for a mission-critical system this is a trade-off with big pay-off, but there is nothing disastrous about a game crashing, it's a bug that can be fixed and patched, where as a rover on Mars is a different story.
In game development you want access to data as fast and wherever you need it, and you want to transform it, Rust makes this difficult to do easily because of how it is obsessed with safety and scopes.
In game development you want the ability to do unsafe things, Rust while it gives you access to the hardware, still wraps you around in safety gear, making things more resistant, and hurting workflow.
In game development you want to write high performance code as fast as possible and want easy access to data, in Rust, you can write high performance code that has lower possibility of run-time bugs due to compiler constraints, but it takes much longer regarding workflow.
C++ is my primary language but I don't subscribe to the insane levels of OOP dogma that's been around for so long now, but I also use and like Rust (although I am no expert of the language) - just don't like it for game development.
Last edited:
By not writing "textbook OOP" C++ that wildly allocates memory all over the codebase while praying that "smart pointers" will handle it for you, and so forth.
Can you explain what you found annoying about this transition? Personally when I work in managed languages I am constantly bothered by my lack of control.
Thanks for posting these. Blow comes off so much less arrogant than I was expecting. HAving mostly seen him at videogame talks and in kind of weird interviews and his social media persona, he comes off incredibly arrogant to me, but in these talks he doesn't come off as arrogant... But pragmatic and reasonable. The videos changed my perspective on him a bit.
I'm nowhere near his level of knowledge but this looks like a case of:
This sounds like a criticism but it's actually good.
The best one will eventually prevail and be used by the majority (but never 100%, because nothing is perfect)
I believe in Jonathan Blow, I love everything he is involved.
I always feel like when people complain about "magic" in C++ they might want to invest a bit more effort into understanding its semantics.
Yes, it absolutely is the case that more "stuff" happens automatically without you spelling it out than in, say, C. That's the whole point of abstraction, after all.
However, compared to many (most I'd say) other languages which offer a similar level of expressiveness it's still rather easy to understand and reliably predict what exactly happens in a C++ program. You just need to invest some time to fully study its value semantics, object lifecycle, and implicit conversion systems. That's pretty much it, as far as "magic" core language semantics which potentially result in runtime behaviour go. Maybe you could also consider some template unfolding/matching "magic", but again, it is absolutely predictable in its manifestation and happens at compile time.
That runtime predictability is IMHO what is missing in many "more dynamic" languages and what is important for game development -- or at least high-end game development to be more precise.
Of course, I'll readily admit that all of this manifests as a severe initial (and sometimes also more long-term) barrier to understanding. And that C++ - especially modern C++ - is a massive language, and in a few cases unreasonably so, is absolutely true. But at least for me, I start to miss quite a few of its features when I work in other languages; even if they are well-designed and suitable for the task at hand, like C# or Rust.
Specifically I had to create a windows credential provider (rfid stuff). I couldn't get the VS debugger to attach to a kernel process on the vm I was developing in. So when my provider spawned multiple child threads tracking down errant pointers causing crash loops was hell with only text out debug statements through windbg.
Other annoyances were:
- Consuming REST services - trivial in managed land (very few lines of code in nodejs, easy frameworks in java) was non trivial in C++ land. In the end I had to settle with a MS based framework which required me to build the whole thing and trouble shoot my environment (does stuff ever build first time without issue?).
- The crazy amount of string formats/wrappers, seems every framework I wanted to add used its own string type. Code examples given on the web using unsafe functions/depreciated code. The official MS documentation leading me around in circles.
His attitude can flip a lot. He doesn't like explaining the same thing multiple times. But he does understand games development.
You specifically said:
How am I supposed to read that as anything but a commentary on gender roles in the USSR in general. If you only intend to comment on their labor participation then just say that, don't throw in deliberately vaguely-worded nonsense that you can backtrack from later.
Just to be clear, I was trying to state Blow's points for people who don't know. I use (a fairly reduced subset of) C++ day to day and don't find its magic gets me too often. I follow Jai out of interest, not because I think C++ desperately needs to be replaced.
I've been messing around in Rust lately and certainly, I hit features I miss from C++ all the time. The most recent is primitive parameters in generics :(. And that language has been around the block for a while.
I don't have a source, but I remember reading that Johnathan Blow called a game named Stephen's Sausage Roll the best puzzle game ever. Just FYI.
Yeah, if the problem is just to crunch something out fast it can get hairy. I mean Rust is basically more legacy free approach to many of the same ideas that C++ had augmented with useful things from other languages along the way. The modern C++ approaches Rust in many ways but is even more complicated and can't often guarantee same level of safety (string_view for example is quite a footgun while the same idea works in Rust). The fact that you have more advanced type system doesn't necessarily mean you don't have fast access to data or nothing to say about memory layouts.
Of course Rust as its current state isn't ideal yet. Last time I did hobby projects with it there wasn't really a way to implement eg. linear allocators properly, not sure if there is ongoing work to solve that. And especially in the beginning Rust is going to be really unforgiving, only time will tell if it's too much.
Last edited:
For what it's worth, I've been following jai since its inception, and I don't think Blow considers C++ magic (as in "doing things I don't understand"). Honestly doesn't seem like a think he would say.
Any discussion on that point without sourcing it seems pointless.
Just to make it explicit for the rest of the thread, from talk#1 @ 19:10:
Primary Goals:
- Friction reduction ("Easy to do the job you're trying to do")
- Joy of programming
- Very fast compilation
- Lack of tedium when expressing self in code
- Pleasant and helpful error messages
- Performance ("Do all the low-level stuff we want to do all the time")
- Simplicity ("Hell of a lot simpler than C++")
- Designed for good programmers
Last edited:
Context for that statement is provided directly in front of that "vaguely-worded nonsense" -- the part you did not quote: "That it is a broad societal rather than domain-specific (or, for that matter, biological) issue can be seen by e.g. comparing the percentage of women in engineering disciplines between ex-communist eastern European countries and western ones."
Maybe I should have repeated that restriction again in the subsequent sentence, but usually in conversations some level of consistent context is assumed.
Oh, I agree, it was just a more general perspective of what I've found with people who express their frustration regarding C++ "magic".
Talking about template parameters, what I really want in C++ is using identifiers as template parameters (as that would allow me to get rid of most xmacros). I brought it up once with someone in the standard committee and it doesn't look like I'll get my wish :P. Then again, it would be mostly obsolete if we ever get metaclasses anyway.
Last edited:
I know this will sound naive but what Media Molecule are doing with Dreams is the future. That and using AI to assist in complex problem solving. I mean proper AI, that you can just ask to do something and it works out the best way and does it way quicker than even the best coder code do.
I don't know who that is lol.
Is this why?
Hope this doesn't blow back in Jonathan's face.
Lol, his name is Jai Courtney, he's an actor who is consistently deemed as terrible and is in terrible movies like Die Hard 5, Terminator Genysis and he plays a character in Suicide Squad called Captain Boomerang.
Hence the Blow back in Jonathan's face line.
Legacy, popularity of an "inferior" standard, stubbornness, inertia, standards that don't really have the capacity to be the single standard for a given area, no clear best one. Take your pick, these are all factors that can prevent a clear choice from being made, and are what lead to the situation in the comic.
For those who don't know what we are talking about C++ is a general purpose language, with it you can program anything and in fact most of the pc software and games are made with it.
C++ is definitively the best choice for performance(unless you go for Assembler but that's basically machine language), but it's a somewhat complex and even for some basic things you need a library.
C#, Java etc are nowhere near as fast as C++.
I'm really curious to see how easy and fast can be his game development focused language, but the biggest problem that C++ contenders have to fight is that C++ has decades of libraries, so even if C++ is complex you will always find a library that can help you, while on a newer language you need to do most of the things yourself.
C++ is definitively the best choice for performance(unless you go for Assembler but that's basically machine language), but it's a somewhat complex and even for some basic things you need a library.
C#, Java etc are nowhere near as fast as C++.
I'm really curious to see how easy and fast can be his game development focused language, but the biggest problem that C++ contenders have to fight is that C++ has decades of libraries, so even if C++ is complex you will always find a library that can help you, while on a newer language you need to do most of the things yourself.
Last edited:
Hmm thinking about it a bit more, you are correct. Fucking windows is the best example for that.
The creation of coding languages for specific niches and especially in gaming isn't exactly new.
I mean, even bigger names do this sort of thing to as the need arises. Naughty Dog made a new language (GOAL) for J&D just so they could get the whole seamless world thing going.
I can see this being useful for beginners or small teams looking to do quick work in short time spans. Could be interesting but unlikely to be something i'll look into.
I mean, even bigger names do this sort of thing to as the need arises. Naughty Dog made a new language (GOAL) for J&D just so they could get the whole seamless world thing going.
I can see this being useful for beginners or small teams looking to do quick work in short time spans. Could be interesting but unlikely to be something i'll look into.
C# is a great language, but it's a lot slower than C++ and has limited memory management. It also tends to be Windows-focused.
Numerous issues, but we could start with something simple, like the most popular implementation having the GIL.
Last edited: