Should we stop writing functions? (C++)
foonathan.net
foonathan.net
- a simpler language that does less
- a less popular language
- a more verbose / restrictive language
If you want people to switch to something better, make something better. Its not enough to simply call out the issues (which is valid, but not useful anymore), and it entirely dismisses that replacement languages like D (small, less active ecosystem) and Rust (hugely verbose and equally bloated) do not actually service the entire range of C++ users.
If all the other compilers and languages were THAT much better, they would be used.
For a lot of apps, Rust has been that replacement, but not for all domains C++ is used in.
Why do people still use telephones when calling via whatsapp, telegram, element or discord exists? Why do people still drive shitty motorbikes when electric cars exist? Why do people still write by hand when you can type instead?
Because the replacement is not a complete replacement and does not actually work better for everyone.
What you are saying is true: I've been writing high performance scientific code and desktop gui apps. I would love to use Rust for my projects, but it just doesn't cut it. The libraries I am using are very mature in C++, but the libraries in Rust to accomplish the same thing are still too immature to consider in my projects.
Also, can you not use these libraries from Rust?
Anything related to HPC and HFT, CUDA, game engines (Unreal/CryEngine/Ogre3D/Godot vs Bevy), Qt/WinUI/MFC/VCL/FireMonkey/wxWidgets/KDE, COM/XPC/Binder, compiler frameworks (Graal/GCC/LLVM).
Yes, many of those could be used from Rust, some of them already are, provided there are bindings, then again it is the classical question if one wants to maintain bindings, or write the application they care about.
We already have C, which is better. It existed before C++. It will exist long after C++ is dead. If you don't over-architect and over-abstract your code you can be more productive in C than in C++ precisely because you avoid the kind of nonsense that this article (and the endless other articles about overcoming C++'s shortcomings and cognitive overheads) talk about. In the time you spend waiting for C++ code to compile, you can instead just write more code in C.
Look at all the articles that get posted here about C++. There was a post a day or two ago about trying to work around C++ compilation times. How long does it take to compile C projects? Not long at all, if you follow standard rules that have been well-known for decades like forbidding #include in header files.
>If all the other compilers and languages were THAT much better, they would be used.
Argument from popularity/Blub paradox. If C++ is so good why does it need to be constantly updated and extended? If you think C++23 is better than C++20, then surely you accept that C++23 can at least potentially be worse than things that already exist.
---
People have an attitude towards C. They think "oh but what about std::vector? I will have to write my own containers". But 90% of the time you will have good enough or even better performance with a few lines of code. Often you know statically the maximum size and can just allocate that at compile time, or you know before filling a collection what the size will be. In C++, most code dealing with vectors that I have seen pre-sizes the vector anyway. That could just be a 'malloc' or a static array. The result is that you compile in debug mode and your program actually works, you can use a debugger, everything Just Works. Programs run at almost full speed in debug mode, and debuggers work. It's glorious.
In C++, you spend a hundred thousand lines of code reimplementing standard library containers anyway because anything more complicated than std::vector is unsuitable for non-trivial programs, and then the only way your program will run is with optimisations on, because your 'zero-cost' abstractions are actually very high cost if you aren't compiling with -O3.
Oh yay I have a big standard library in C++. Thanks for saving me from having to use the old ugly BSD sockets API to do networking. I can use a nice modern networking API in C++. Oh... it doesn't exist? Oh well, I'll use a nice modern filesystem API in C++. Oh, it only was added recently and lacks basic functionality? Oh well, I guess in return for learning the 50 different ways of initialising a variable, and all the subtleties of argument-dependent lookup and glprxvalues and template deduction guides and template⟨T=std::enable_if⟨std::is_void⟨void_t⟨int⟩⟩::type⟩::value⟩ I will get the benefit of using a high-quality standard library that contains std::regex and std::random and std::chrono and std::thread and std::vector<bool> and std::unordered_map and all the other VERY high quality standard library facilities provided by the C++ standard library!
Oh they're all crap. Pity.
You maybe can write small program easier on C as long as you can fit everything in someone's head. And that's why all complex low level programs are written in C++ and not C. (With the Linux Kernel being an exception, because of its stupborn maintainer)
And what about the drawbacks mentioned on the article? 99% of programmers don't need to care about them. The reasons such article exists is precisely because some people love C++ and like to play with their language.
---
> If C++ is so good why does it need to be constantly updated and extended?
Uh? To adapt and evolve to a changing environment. Instead of stagnating and becoming irrelevant. If it is so great, why is the iPhone constantly release new models? Why do every maintained software release new versions? And that's why you should use C++ instead of C.
I explained why it doesn't matter that "C lacks basic primitives...". C has the basic primitives. It lacks generalist containers, for example. But most of those 'generalist containers' are, as I already explained, just not very good. For example, take this post:
https://zeux.io/2023/06/30/efficient-jagged-arrays/
What do we see here? The guy starts off with basically 'normal C++' and what he has at the end if essentially something you could write in C. When you optimise C++, you end up getting C. He has two std::vectors, but they're pre-sized so they're essentially the same as writing 'unsigned offsets = calloc(vertex_count + 1, sizeof offsets);'. This is not any kind of significant boilerplate.
The low-level details are something you NEED to concentrate on. If you don't care about the low-level details, then you shouldn't use C or C++. That's the point of low-level languages: control over low-level details.
All complex low-level programs are written in C. C is used in every production operating system kernel. C is the language in which systemd is written. It's the language in which pretty much every production web server or network server is written in. It's the language used by pretty much everyone interfacing with hardware.
The Linux kernel isn't an exception. It is the typical case.
>And what about the drawbacks mentioned on the article? 99% of programmers don't need to care about them.
The drawbacks mentioned in the article are major issues. They regularly bite C++ programmers, and are always lurking. You can't just say '99% of programmers don't need to care about them'. The whole point of namespaces in C++ was that you didn't need to care about name clashes. The result of argument-dependent lookup is that instead of getting a linker error on a name clash, like you do in a sane language, you instead can get random weird functions being called, silently.
>Uh? To adapt and evolve to a changing environment. Instead of stagnating and becoming irrelevant.
You've missed the point. You say C++ is so wonderful, but it's one of the languages that today is changing the most in idiom and specification. Since C++11, we're constantly being told about the new 'proper' way of doing things in C++. They're adding new features then going 'oh actually that was a bad idea' in every new revision. If it were so great, why would it be changing so much?
None of its changes are to 'evolve to a changing environment'. In C++11? Sure, they added threads, they added atomics. Pretty hard to argue that it wasn't an evolution to adapt to the new multithreading world. But today C++ still doesn't even have a networking API standardised. It still doesn't have a decent filesystem API. It can't even adapt to the world that existed when it was FIRST standardised, back in 1996. How is anything they're doing adapting to the world of today?
---
Really none of what you said addressed anything I actually said.
You could make that argument about almost any technology (certainly in the IT world). C23 is out this year which will make it the 6th "official" version - obviously a much more slowly evolving language than most, but nonetheless still being fairly regularly "updated and extended".
There's plenty of things I don't care for in C++, but I can (and do) choose not to use those parts. The few times recently I had to work with pure "C" code felt, well, decidedly primitive - but there was no choice but to grin and bear it. Which works great for some contexts, but I'm not sure I can imagine tolerating it for large-scale application development.
The problem is that it's a language that doesn't need to change much. It doesn't really need a committee. People that recognise this don't bother getting involved. That leaves it vulnerable to entryists.
You cannot pick and choose which C++ features to use in practice if you use any kind of library including the standard library.
Millions of professional C++ developers work on large scale mission critical C++ code that control everything from airlines to Fortune 500 companies to cars to nuclear reactors to military hardware. And new C++ projects are started every single day.
Most developers have zero clue on what it takes for a programming language to be successful in the real world. And they have zero clue on how much $ and time it would take to rewrite the billions of lines of C++ code out there that runs the world.
The closest overall in tooling IMO is Rust, but it's currently missing a lot of the above, and very many C++ abstractions cannot be elegantly expressed in Rust due to limitations in generics, const functions, the orphan rule, and a lack of C++-like functor structs.
What? Exactly what led you to think that nonsense? A random blog post of someone misusing lambda expressions?
There are plenty of things in C++ to criticize, but a) you are not actually criticising anything at all and instead spewed just noise b) misusing a feature is not a problem of C++.
The author of the blog post is in fact one of the authors of fmtlib.
Getting libraries to play nicely with ADL is actually a huge pain in the ass and in fact the three major C++ compilers don't agree with one another on how to actually perform name lookup in a large number of cases.
There's a huge number of business process where execution time is almost a non-issue. As long as the thing runs every day/week/month/quarter no one will give a crap. Development time, supportabilty, and onboarding time for new developers tend to matter way more for those sorts of processes. For things your shipping out to end users? Optimization and speed tend to matter enough to justify the development cost, but there's an absolute ton of internal business processes, and even internal applications, where optimizing it would cost way more than just running with what the dev team already knows.
C++ is generally used by millions of people to whom execution time matters quite a bit. The language is literally designed to force the programmer to decide how to balance CPU cycles with language abstractions and features. It's already hard to debug compared with a lot of other languages (which are generally slower).
The fact that you (and GGP) rarely encounter situations where run time is critical is a limitation on your experience, not a universal law that applies everywhere.
So, context is really important here.
I think that is a matter of some debate.
>> Efficiency has been a major design goal for C++ from the beginning, also the principle of “zero overhead” for any feature that is not used in a program. It has been a guiding principle from the earliest days of C++ that “you don’t pay for what you don’t use”.
https://www.think-cell.com/en/career/devblog/should-we-stop-...
Haters don't waste time trying to understand what is before them. They just see a superficial opportunity to unleash hate and just go for it, logic be damned.