I Tried JAI, Can It Replace C++?! (Programming Language) 

🤓

Oh my, I always thought it was neighter, pronounced it as Jai...

Jai ≠ Jai ?

You mean Jay, not J?

@@peezieforestem5078 He means Jai, not Jai or Jai. Just Jai.

What lol, so you have a clip of that. Sounds funny as hell

@@Cakez77 ru-vid.comUgkxiIlkFWoCwX2jzIo-4F3I53Ae0hBuNfQ8?si=z3fi_J-vPPiiDUxl

Links are shadowbanned by RU-vid btw

@@tech6hutch Usually they have a thing in the dashboard that allows the creator to show links if they so choose to let them through. Unless if they removed the option to verify links, it'd be up to him whether or not he saw it.

idk, but there were "2 replies" and I only saw one, hence why I said something (since I'd also like to see this lol)@@anthonyewell3470

Huh interesting, always thought it was Jai. Wonder what happens if he renames the language lol

Omg thank you. I looked everywhere for how to pronounce it and finally was like, there's no way. He has to be doing this on purpose! Drives me crazy. Just say it once! >.< Is it Jai or Jai???

"our language" **marxism intensifies**

Or simply break ; and it breaks out of the loop

@@andreffrosa Hey that's just goto designed to appease people told never to use goto.

That was my last use of goto. But since C++11's lambda feature, I never touched goto again.

Came here to say this. Also C is not problematic because you can't break out of two for loops. And having to include external libraries to write a game in a general purpose programming language is also not a problem (what might be problematic is that dependency management is not part of C itself).

And dont forget the clean code recommendations that you need a single file just for every class, no matter how small (even enums) and that you need to include the header as the first include in the source (to make sure headers are self contained) which eliminate the ability to use precompiled headers. Also so many things could be implemented if we had whole file compilation and made the linker to do a little more (like the Objective-C linker does).

Partial template specialisation is not much slower than just a new class/procedure. It's an instance when you use it, for each compilation unit, that the template is instantiated that takes the most time. Especially stuff like variadic templates often call more templates, cascading into an enormous amount of generated code, that may well be duplicated in another compilation unit. Why you should just ignore people that say you should use make_unique and just use new, for instance. And forward declare all the things. I recommend actually profiling the build to find out what is taking all that time. You can do that in clang with -ftime-trace and dropping the json in some flame graphing program.

This is a fair point, but build systems and especially CMake add another layer of problems on top of that. CMake behaviour is often not intuitive, so It's very easy to accidentaly introduce spurious dependencies or duplicate jobs. In my experience, the compilation model of C++, combined with the way it's configured through CMake is where everything breaks. e.g. build fails because you included a header file from different target - oops, so you add the cpp file into your target and suddenly that file is built twice (even though you'd think CMake will build it once and cache it). Or, you already know about this problem, so you link the other target with your target instead, but now your target can't start building cpp files until the other completes, because CMake adds the dependency edge on full build, not on link step. I've found a ton of issues like this in the build at my previous job, 60% of the build time was basically wasted on nonsense like this.

Noted, glad you like it. I actually have one more comming

Yeah, I was surprised too, but I also saw the SDL2 lib, so that does exist

its using glfw3

Hey that's great to hear, glad I could inspire you

Stoked for this!

Subbed, waiting for that jai vid

@@MenkoDany thanks guys!

Thanks, would really like to see those metaprogramming features your talking about. I have never used anything like it even in c++

@@Cakez77 the compiler is an event loop that can be hooked into to control everything the compiler is doing.

Is there a way for non-personality non-game devs to get access? I'm an ML person who is frustrated with some tools and curious if/how jai solves them.

@@VACatholic If you haven't already considered it, have you tried Odin? It doesn't have the metaprogramming, but it has many similarities and shared philosophies. Unfortunately, JB has said that the public interest in Jai is far too strong at the moment (they're getting spammed with requests), so we're discouraged from sharing how to get on right now. On another note, JB has recently said his game is getting closer to release, which I think means between 6-24 months, and if he sticks to the original plan, the language will be made mature and open relatively soon after the game is in production with low maintenance.

@@Muskar2 Fair enough. I have a day job, so it'd just be a hobby anyway.

Don't worry, it will be out in 3023 👍

@@Cakez77 how did you get into the beta? Just emailed jblow? I would email him but I'm scared of rejection since I work in the financial sector now and the last game I worked on was before 2020

Okay looks like that is what I'm doing next

The proper way of learning yep also more fun

I don't know about the higher order functions tho, I'm not that involved yet

It's just how I say it lol, most be my German seeping in

indeed.

Yeah you can use go-to, bot I found that it makes the code harder to read unfortunately

you can absolutely do that, and its common practice when working on kernels or drivers.

I will try out Odin soon, but it sounds really good man

Yeah, thanks for the suggestion. I will try out Zig and Odin too and then make a comparison

Seems quite coupled with the update loop of the engine tho. But maybe that's good?

Haven't tried the other two, but Jai was really good

My bet's on Zig, and it's not even close imo. I think the other two are *waaay* farther away from hitting any sort of critical mass.

What makes Zig so good?

@@Cakez77​ @delphicdescant i guess so just for the case that it comes with c/c++ compilers and build system so like we already have big code base that cant be rewritten so they might try the build system as a start then might start using zig along with their prev code base thats where Zig even comes on top of Rust its just its not yet prod ready v1

​@@Cakez77its comptime Feature and low-level-support (combined with the ability to use C Code)

Yeah me too, I think it's building it as a unity build internally. I get similar build times with c++ where using unity build

There are 3 main reasons for this: 1) comparing to c++, jai is designed much easier to analyze (especially metaprogramming part) 2) it doesn't do a lot of crazy optimizations stuff, that gcc does and instead gives user easy way to optimize their code with "modern" techniques like allocator switching. 3) there are no incremental building by default

I'm actually watching his videos on that yeah, been a lot of fun

What videos are these? Would love to see more.

Would love a link to his channel. I found channel by name randy who is a game dev but his videos are shorts and more show case oriented than programming.

Would the intro trigger him? :O

​@@Cakez77He advocates that C++ is a really terrible language, but it does somethings right that most other languages fail to do, so that's why we use it in games. This comes in complete contrast with your opening statement lol, and then you go on to say that it has some problems and the first thing you mention is that yoi can't break out of 2 for loops, and i laughed because he would consider this reason such a trivial not even worth mentioning shortcoming of the language, that's why I said it lol. You may want to watch his 2018 talk "Jon Blow's Design decisions on creating Jai a new language for game programmers", I find that interesting

You have to ask Jonathan Blow for access. Write him an E-Mail with who you are and what you would like to do

@@Cakez77 Is that how you got access? My impression is that most get access by getting a key from someone already in the beta.

Second time I hear this, will do yeah. Sounds interesting

We will see, seems like there is Zig as well

I will try out Odin soon, I wonder if it's good

​​​@@Cakez77 Odin took a lot of inspirations from jai, the first thing that you would notice is how similar the syntax is. Also Odin is used in production, there's this company called gangafx which Bill (odin creator) work in, wrote its flagship real-time volumetric fluid simulation tool in Odin, EmberGen, they also WIP two new software, GeoGen for terrain generation and LiguiGen for liquid simulation. EmberGen used widely in Hollywood and some game development studies use it too.

The goal is to support mobile, but as it stands I don't think it's supported

i also thought about this

I have never heard of it, 🤔

@@Cakez77 Odin is more ergonomic than Zig and also has inbuilt features that makes it easier to do gaming, VFX type development. Odin has the best error handling when compared to Zig, Go(and even something fancy as Rust) according to Rickard Andersson who was so impressed by the language and how the features work together that he made a whole Odin programming playlist on youtube.

Sounds interesting, what makes it so good compared to c++? Maybe I should really try it out

@@Cakez77 the c/c++ interop alone makes it one of the best languages for people coming over from the (C)lan

i agree... with zigs interoperable nature with c and how EASY you can port in the raylib c library... Honestly why bother with learning another language...

I agree. With Zig out to replace C, Odin aimed at game dev, and Nim being so amazingly easy, fast, and compiling to C/C++, Objective-C, and JavaScript, Jai is late to the party and really has to stand out to gain traction.

There's no difference in Jai. You can use any C library and even generate bindings automatically @@ev3rybodygets177

Oh, it's called RenderDoc, it allows you to inspect graphics applications

Thank you @@Cakez77

It's just a .zip file to download. As far as I know there is no "cloning"

​@@Cakez77I guess they meant to use the command `git clone --recursive`

oh lol it was your code? Well yeah it was very helpful to get started. In my opinion it should be part of the examples

Because that's how I say it? Dunno

You can implement your Option/Maybe using a struct with an enum or bool and a polymorphic variable inside, and some functions if you don't want to check it manually.

Lol

And those things would be? Never tried golang

Golang is garbage collected. Just stop

Ja-eye

Lol

I think the problem of rust is the borrow checker but I have never used it so I have to try it first to know for sure

@@Cakez77 Garbage collection vs borrow checker.

@@TheBlackMaster Its not GC vs BC. I actually have a language idea in my mind that does not do borrow checking, but neither GC and in memory safety its not as good as rust, but handles 99% of cases (and only error in remaining 1% if you are noob) - all while its much-much more simple and let a lot of valid and safe programs through that rust does not. For me rust is so much productivity pain for useless gain (I do not make those issues anyways in modern cpp) that I rather use modern C++ and code faster...

Well, the developer's own machine compiles 250,000 lines per second and he wants to hit 1 million once it is more feature stable.

Lol 😂 I mean for me it is. The good thing is that the library is maintained by Jai and not a third party to that you have to add manually. That's mostly my point

That is not the main point of the language. The metaprogramming and compile-time execution features plus a bunch of smaller design decisions are the big selling points. Honestly, I think the in-program build system is a huge deal.

KEKW

You can jai deez nutz bro

But they could come back if the language is good when it releases, you never know

@@Cakez77 I think it would have to be a lot better than what they are currently using to entice them to make the switch. and honestly from what I've seen it's not better than either of the two I mentioned.

Assuming Jon Blow even wants such attention like that

If you've ever followed *any* of Jon's projects you'll know he doesn't rush things out to beat the competition. That's not his strategy and never will be. He doesn't care about winning the short-term race, he cares about making the best goddamn language he can, and so far, he's doing a pretty fine job of it IMO. But yeah.. it totally sucks that not everybody has easy access to it, and I can understand how it feels dead if you're not in the beta and getting the weekly release notes of the insane amount of stuff being improved and changed. He's also the CEO of a company that's building quite a few other projects right now (Braid: Anniversary Edition just got a release date announced, and he constantly streams his Sokoban game's development on twitch, plus other, yet-unannounced projects). Building great things takes time.

@@dandymcgee zig and Odin aren't being 'rushed' either, yet those are two great products that are on par or better than jai that people can use right now. the reason jai isn't out now is because Jon is an elitist and his ego can't take getting criticisms, or suggestions from people he considers to be beneath him.

Confirmed

yeah it could be lol

As one of the participants said (maybe it was tsoding, but not sure): "instead of ending the closed beta, it will just grow enough so everyone who interested will have access"

@@cyanmargh I was interested, 5 fkng years ago, never got access. Now I'm all about Rust anyway so who cares

It's definitely switch up yeah. Takes time getting used to

@@Cakez77 It's not even that it's a readjustment - it has been proven that it is actually worse for readability to have types right aligned instead of left like C. Why? Because your brain can parse the types of whole blocks of variables much faster that way, without even having to read them. Here's a basic example with just some primitive types: float myFirstFloat float mySecondFloat float myThirdFloat int someInteger int breaksNamingConvention bool someCondition string someText Compared to: myFirstFloat float mySecondFloat float myThirdFloat float someInteger int breaksNamingConvention int someCondition bool someText string As you may be able to tell, finding a specific variable of a specific type is much harder this way. In the case of primitives specifically, there's actually a funny thing where they are all (mostly) have different name character lengths, making them *even easier* to parse when they are aligned in a block like one would do in C: int (3) bool (4) float (5) string (6) That's why, as someone who really needs and wants this code readability as I have poor eyesight, this "trend" in many modern languages to change this is genuinely extremely frustrating. The over-use of type inference is similarly annoying, as understanding what a block of code does and what types it has gets much harder when they aren't explicitly stated anywhere - I mean, try reading a python script and understanding what any variable's type is. Thanks for coming to my TED Talk.

@@nolram That's a lot of nonsense. The type is rarely important, and if you can't infer it from the context, your code probably sucks. Using Python, a dynamically typed language, as an argument for type inference in a statically typed one shows a clear lack of experience/knowledge.

@@Giga4ever Well, a dynamically typed language is just an extreme example to visualise the point. Types in statically typed languages ALWAYS matter, especially for readability of code. Knowing what data your code operates on is key - there's very few more important things. And "inferring from context" not only takes more time than just... seeing it, it also leads to less awareness, and especially in larger codebases you can't rely on knowing the context. Using explicit types makes code more readable and faster to parse and is superior in almost every single scenario. You read your code more than you write it.

That's not readability being worse it's just how used to it you are.

Gonna try Zig and Rust to see for myself, but I have heard bad things about rust compilation times.

​@@Cakez77 The people who are complaining about the compile times are the people who pull in hundreds of dependencies. A game written in pure OpenGL and with the Windiows-API compiles in 0.5 seconds for me.

I think Rust and Jai aim for different things. Rust ensures program correctness at the expense of compile-time. Jai ensures fast compile-time at the expense of program correctness. One thing that is very powerful about Jai is it has powerful reflection capabilities - you can write Jai code that dynamically reads the output of the parser and makes adjustments to it before the code is compiled. This means you can very easily write code-generators in Jai for generating boilerplate for you. Rust and C++ at the moment don't support reflection like that.

​​@oscarsmith-jones4108 Wouldn't it be possible in Jai to write static compile time checks that enforce safety in a targeted way so that new devs don't break project-specific rules? For the really hard problems this means static checks and project-specific rules can be enforced that even rust cannot do but with fast compile times.