Programming-language

[tukkek]

Since the original Project M is written in C++, Rust has been suggested as an option. Others have been mentioned as well.

Here's an initial discussion on Discord: https://discord.com/channels/737206408482914387/737206408948220057/1282388486925717574

[kblaschke]

I guess with the goal of having maximum portability and good speed, the choice of the programming language really just boils down to C, C++ or Rust, as any other language is either highly platform-dependent or not portable enough. I'm a C++ guy and don't really have much experience with Rust, but I'll try to make a list of pros and cons here for each choice:

C++

Pros:

• Well-established language with a huge ecosystem of available libraries and frameworks
• Object-oriented
• Fast
• Highly portable onto basically any platform

Cons:

• Wrong memory management can easily cause crashes and security issues
• Compiler differences on each platform require some special care
• Complex build process and dependency management

Rust

Pros:

• Creates highly secure appplications due to compile-time memory safety checks
• A tad bit faster than C++
• Highly portable onto basically any platform
• Easy dependency management via Cargo

Cons:

• Relatively new language, thus still a relatively small user base and ecosystem
• Procedural language (with a bit of additional syntactic sugar though)
• Slow compile times

There are probably plenty more arguments to be added to the above list.

[tukkek]

Crashes were discussed as one of the reasons the current Project M failed to gain traction as an embedded visualizer in music-players. If Rust's safety can prevent this from happening to a relevant amount (which I imagine it should), then it would be a major pro-argument.

(I say that, however as someone who can't write Rust code and hasn't contributed any code to Project M).

A potentially major downside too, in terms of the eco-system, as mentioned is the lack of options when it comes to fully-featured, mature and stable DSP Rust libraries as discussed in #2.

While I have been able to find multiple such Rust libraries, none of them seem to be fully-featured as is something like Python's librosa.

[kblaschke]

The crashes had multiple reasons, many of them were results of passing bad data to OpenGL, from which Rust won't save you. Same goes for any crashes which were the result of unhandled exceptions, which I guess in Rust would also cause a process to terminate. Rust is safer than C/C++, but definitely not a tool that magically fixes all programming errors.

I've fixed all known crashes from projectM 3.1 over the past few years, and most could have been prevented with slightly better code quality and programming techniques. While C++ is definitely prone to memory issues, certain patterns like RAII, properly using smart pointers and having well-known ownership concepts will reduce that problem to almost zero. Dealing with user data is probably the main reason of concern, so extra care has to be taken when reading/parsing external data.

I'm a C++ dev, working professionally with this language for over a decade now, and I've seen good and bad code. With Rust, I've only had a few points of contact, and never written anything in it on my own, so I can just relay what I'm reading and hearing, plus a tiny bit of my own experience. That said, if the majority says the library should be written in Rust, I'd have to trust the dev(s) who will then implement it that they employ similar good practices

[revmischa]

I think the dependency management is infinitely easier with rust, as is cross-compiling. And dependency management. We don't have a great way to pull in deps for our project right now.

[tukkek]

Even though I'm not a Rust user, cross-compilation is a major feature for me in general.The two technology stacks I use that offer it save me dozens, if not hundreds of hours of time. Not to mention not needing to set-up and maintain multiple build-environments... 😭

Also, Rust has a one-liner for Web Assembly builds: wasm-pack build --target web. #6 mentions Web GL, so being able to share code between engine and web sounds like it could be an equally beneficial time-saver (again: no personal experience, just seems like another good opportunity for synergy).

[kblaschke]

Yeah, the Rust package management is definitely a big plus.

If we can actually find two or three experienced Rust devs which can write a larger library with all the features needed, that'd be great! On the other hand, if there's no developer who can actually write the actual code, then deciding to use Rust might just sound nice, but won't give us any useful result in the end.

And yeah, good C++ devs are probably almost as rare. 🙄 But writing such a library in C# or JavaScript is probably not an option...