> We know the result isn’t idiomatic Rust, and there’s a lot that can be simplified once we’re comfortable retiring the C++ pipeline. That cleanup will come in time.
Correct me if I’m wrong since I don’t know these two languages, but like some other languages, doing things the idiomatic way could be dramatically different. Is “cleanup” doing a lot of heavy lifting here? Could that also mean another complete rewrite from scratch?
A startup switching languages after years of development is usually a big red flag. “We are rewriting it in X” posts always preceded “We are shutting down”. I wish them luck though!
A mitigating factor in this case is the C++ and Rust are both multi-paradigm languages. You can quite reasonably represent most C++ patterns in Rust, even if it might not be quite how you'd write Rust in the first place.
In addition, C++ and Rust are very, very similar languages. Almost everything in C++ translates easily, including low level stuff and template shenanigans. There's only a few "oh shit there's no analog" things, like template specialization or virtual inheritance.
Out of all the languages rust takes inspiration from, id rank C++ at the top of the list.
Strong disagree. Rust copied C++ syntax to avoid looking weird to C++ programmers, but the similarity is skin deep. C can be tamed, because it's mostly a subset of Rust, but C++ idioms are a death from papercuts.
OOP, weakly-typed templates, and mutable aliasing create impedance mismatch in almost every C++ API.
Rust doesn't have data inheritance, and what looks like interface inheritance is merely extra requirements in a flat list of traits, so subclassing won't behave like C++ APIs expect. When you translate a class hierarchy to Rust, it needs lots of crutches which make it weird, boilerplatey, and tedious to use. There's no good recipe for OOP hierarchy in Rust, because the idioms are so different. The mismatch feels like writing an ORM.
For some C++ APIs mutability and circular references can be a pain too. Rust works well with DAG data structures and clear mostly-immutable data flow. Objects with some "parent" pointer are common in C++, but Rust sees them as potentially dangling, with shared mutable state, and requires much heavier control of them. It can be done, but it's ugly. Idiomatic Rust designs go to great lengths to avoid it unless necessary, but C++ APIs can have the extra pointers "for convenience".
There's a reason why Rust doesn't have typical GUI libraries - an arbitrary web of references between widgets and event handlers make it ugly in Rust, and that's on top of a view class inheritance.
C++ templates sit very uncomfortably between Rust's macros (duck typed) and Rust's generics (strictly typed at point of declaration).
C++ templates almost always are a mix of types they're attached to and some duck-typing in their expansion.
Rust's generics do not allow any duck typing at all. This makes translation of even a tiny bit clever C++ templates a chore. There's no specialization. No way to deal with SFINAE and such.
Rust macros have flexibility for all the syntax shenanigans (and even similarly bad errors at instantiation time), but macros can't see any types. Idiomatic Rust has very deliberate division between traits (usually much simpler and smaller in scope), macros and proc macros/derives. Splitting C++ templates like that can be a major redesign.
As someone who's worked in both C++ and Rust, they're deeply similar languages. There are far more exotic languages out there. APL. Erlang. Forth.
In the grander scheme of things, Java and C# are literally the same language, Rust and C++ are twins, C is their dad, Forth is the owl staring through the dining room window, and Erlang is an alien spaceship passing over the house.
You can write "C with classes" in the "C/C++/Rust" language, and have users of all three tell you you're doing it wrong. The commonalities between them are mostly necessities of systems programming, FFI, and LLVM-style optimizations, but they all try to get there in different ways.
> There are far more exotic languages out there. APL. Erlang. Forth
Early Rust started as an Erlang clone, with task supervision trees and all (panics, lock poisoning, and the defunct UnwindSafe trait are vestigial features from that experiment).
Rust has more in common with Ocaml than C or C++. It's not a C family language, it just took C-like syntax to avoid the stigma of being "exotic" like the languages that inspired it.
Rust did copy move semantics from C++, but even that ended up being different enough to be incompatible with basic C++ design patterns. Rust chose to have only trivially movable types and guarantee absence of move constructors, so it can't even safely pass a C++ std::string.
> It's not a C family language, it just took C-like syntax to avoid the stigma of being "exotic" like the languages that inspired it.
I agree that Rust has very conventional C-like syntax, but I'd go further and say it has fairly conventional C-like semantics too, at least in its current iteration. I think you could safely duck type Rust itself as a C-like. :)
Strong agree that Rust has drawn great inspiration from the ML family by way of OCaml. Rust is a C-like with ML characteristics, which are its greatest quality, imho. On the other hand, I don't think there's much trace of Erlang left in Rust -- there's probably more Erlang in Go than Rust, and Go is squarely a C-like.
I think it's fair to say that Rust started out more exotic than it is now, but perhaps all that safe, C-like syntax has led to convergent evolution towards other C-likes?
> Rust did copy move semantics from C++, but even that ended up being different enough to be incompatible with basic C++ design patterns. Rust chose to have only trivially movable types and guarantee absence of move constructors, so it can't even safely pass a C++ std::string.
Right, but compare all of that to Erlang's actor-based message passing, where mandatory immutability averts the need to move or lock anything at all, and I think it's clear why one would group Rust with C++, and not Erlang.
I disagree. You can't even create simple C++ inheritance examples because you don't have data inheritance. So basically classical OOP is out of the window.
That's the biggest difference to C++ and most mainstream languages, you simply can't do OOP (which in my books is a good thing) and it forces you more towards traits and composition.
This is the famous trap that Joel on Software talked about in a blog post long time ago.
If you do a rewrite you essentially put everything else on halt while rewriting.
If you keep doing feature dev on the old while another "tiger team" is doing the rewrite port then these two teams are essentially in a race against each other and the port will likely never catch up. (Depending on relative velocities)
Maybe they think that they can to this LLM assisted tools in a big bang approach quickly and then continue from there without spending too much time on it.
I’ve been part of at least 2 successful rewrites. I think that Joel’s post is too often taken as gospel. Sometimes a rewrite is the best way forward.
Moving Ladybird from C++ to a safer more modern language is a real differentiator vs other browsers, and will probably pay dividends. Doing it now is better than doing it once ladybird is fully established.
One last point about rewrites: you can look at any industry disruptor as essentially a team that did a from-scratch rewrite of their competitors and won because the rewrite was better.
> I’ve been part of at least 2 successful rewrites. I think that Joel’s post is too often taken as gospel. Sometimes a rewrite is the best way forward.
HN nerd-snipe alert! OK, you got me good. Can you share some battle stories? I have also been part of rewrites in my career, but my experience is mixed. I'm not here to simple brush away your experience; I want to know more about why you think (in retrospective) it was a good idea and why it was successful.
I can recall recently, listening to an Oxide and Friends podcast where they spent 30 minutes dumping all over "Agile Dev", only to have a very senior, hands-on guy join from AWS and absolutely deliver the smack down. (Personally, I have no positive experiences with Agile Dev, but this guy really stunned the room into silence.) The best part: The Oxide crew immediately recognized the positive experence and backed off the give this guy the space he needed to tell and interesting story. (Hats off the Ox crew for doing that... even if I, personally, have zero love for Agile Dev.)
> Modern C++ pretty much solves the safety issues.
I always wonder how can one come to such a conclusion. Modern C++ has no way to enforce relationship between two objects in memory and the shared xor mutable rule, which means it can't even do the basic checks that are the foundation of Rust's safety features.
Of course, this statement is also trivially debunked by the reality of any major C++ program with complexity and attack surface of something like a browser. Modern C++ certainly didn't save Chrome from CVEs. They ban a bunch of C++ features, enforce the rule of two, and do a bunch of hardening and fuzzing on top of it and they still don't get spared from safety issues.
FWIW Chrome includes third party libraries like freetype and lots of bugs are in javascript. I imagine defensive checks in javascript will be controversial since performance of javascript is controlled by webdev, not by browser.
Note that Chrome is replacing[1] FreeType with Skrifa[2], which is a Rust-based library that can handle a lot of the things FreeType is being used for in Chrome. A lot of Chrome's dependencies are being rewritten in Rust.
This article is not really specifying if Rust is compared against "Modern C++" or "Old School C++". The only thing we can assume is that at least part of it is "Google C++".
Maybe we would see similar effects adopting Modern C++. Maybe not. The article doesn't tell us.
This is a very shallow, very boring criticism. I doubt it will resonate. Modern C++ does not solve the safety issues, it has plenty of brand new footguns like string_view. Who cares if Go is better than Rust? Feel free to write Go, no one cares.
"mut and fn very annoying to read" like okay lol who cares? What should anyone take from your post other than that you aren't that into Rust?
> Go is a better Rust. Rust is an ugly version of C++ with longer compile times and a band of zealous missionaries.
Eh. There's a lot I like about Go. I adore its compilation speed and the focus on language simplicity. But its got plenty of drawbacks too. Default nullability is a huge mistake. And result types (zig, swift, rust) are way better than go's error handling. Sum types in general are missing from Go, and once you start using them its so hard to go back. Go also doesn't have anywhere near as good interop with native code. Mixing C (or any other LLVM langauge) with rust is easy and feels great. You even get LTO across the language barrier.
The big thing I'm growing to dislike about rust is how many transitive dependencies a lot of projects end up pulling in. Its very easy to end up with projects that take a million years to compile & produce huge binaries. Not because they do a lot but simply because everything depends on everything, and the dependency tree takes a long time to bottom out. I don't know what the right answer is. It feels more like a cultural problem than a language / ecosystem problem. But I wish rust projects felt as lightweight and small as most C projects I've worked with. I'm doing some work with the stalwart email server at the moment (written in rust). Stalwart is a relatively new, well written email server. But it somehow pulls in 893 transitive dependencies! I'm not even joking. Compiling stalwart takes about 20 minutes, and the compilation process generates several gigabytes of intermediate build assets. What a mess.
20 minutes! What hardware is this on? I've worked on Rust projects with similar numbers of dependencies where the compile time (for a clean release build) was 2-4 minutes (on a MacBook M1 Pro)
UPDATE: tried compiling stalwart on my machine, and it took 14 minutes, with a really weird timing profile:
- 99% of the ~700 crates were done compiling in about a minute or 2
- RocksDB (a C++ dependency) was 2 minutes by itself
- And then it took 10 minutes (ten!) just for the final binary at the end.
That's not normal for Rust code at all. Even large ones like Servo or Rustc or Zed.
UPDATE2: turns out they have LTO enabled by default. Disabling that brings the compile time down to 7 minutes. But that's still really unexpectedly slow.
> But it’s still too many dependencies and too slow.
I definitely agree that it's too slow. I just don't think the cause is "too many dependencies" because I've compiled Rust codebases with twice as many dependencies in half the time!
It seems to produce a 94MB binary. So it may be partly that there are some very big dependencies. But the amount of compilation time that ends up in the top-level crate (even with LTO disabled) also makes me feel like this must be triggering a compiler bug. Either that or using far too many generics.
I agree it's probably monomorphization (speculation without looking at it). Generic function parameters might be the root cause, but number of dependencies is a combinatorial multiplier.
I've hit compiler bugs that behave this way. Here's one from an LLVM upgrade [1]. The test case I discovered apparently took over 20 minutes to compile, up from 26 seconds on stable! Their value tracking algorithm was accidentally quadratic.
Using interface as it was designed to be used offers first-class sum types. Although not all interface use equates to sum types.
But they're not tagged unions. I expect that is still where your confusion lies. Tagged unions and sum types are not equivalent. Tagged unions are a subset of sum types.
This may be a bit too pedantic, but I consider interface {} to be a way to do polymorphism via type classes. Interface defines an open class of types which implement some interface.
Sum types are a type definition defining something as A or B. Not “anything that quacks like a duck”. But concretely “one of this or one of that”. This enables different syntax, like the match expression to be used, in which you exhaustively list all the variants. The compiler doesn’t need to heap allocate enums because it knows the maximum size of a single value. The compiler and programmer can take advantage of the knowledge that there’s a closed set of values the type can hold. It’s not an open type class.
Result and Option are quite beautiful as sum types. But they’re horrible as type classes. Try implementing them using interface{} in Go. It’s not the same.
But can also define a closed set of types, perfectly satisfying "sum types".
> This enables different syntax, like the match expression to be used, in which you exhaustively list all the variants.
Go does not provide this out of the box, but that is not prerequisite for sum types. The mathematical theory says nothing about "the compiler must produce an error if the user doesn't match call cases". There is sufficient information provided by the sum types if you wish to add this to your build chain yourself, of course.
By that definition a void* pointer in C is a sum type. By that definition assembly has sum types.
This argument feels like the “we have sum types at home” meme. Ergonomics matter.
I write a lot of rust. Rust has traits which are similar to Go’s interfaces. But the features aren’t the same, and I use enum all the time. (I also use trait all the time, but I use trait and enum for different things).
> By that definition a void* pointer in C is a sum type.
No. That doesn't make any sense. void* is essentially equivalent to any in Go, which isn't sum types either.
You can construct sum types in C by combining structs, enums, and unions, but it is not an out of the box feature like in Go. Sum types are a first-class citizen in Go.
> Ergonomics matter.
Math doesn't care about ergonomics. You might care about ergonomics, but logically when talking about those ergonomics you'd call those ergonomics by name, not by some unrelated thing from type theory.
> You can construct sum types in C by combining structs and unions, but it is not an out of the box feature like in Go. Sum types are a first-class citizen in Go.
Maybe I'm misunderstanding what you mean. Can you give me an example? How would you write a Result type in Go? (Result is defined as either Ok(val) or Err(err))
enum result_type {
OK,
ERROR
};
union result_value {
int value;
char *error_message;
};
struct result {
enum result_type tag;
union result_value value;
};
This is not technically closed, but does offer a close enough approximation. Again, not a first-class feature, so there is no expectation if it being true sum types.
Go:
type Result interface {
isResult()
}
type OK[T any] struct {
Value T
}
func (OK[T]) isResult() {}
type Error struct {
Message string
}
func (Error) isResult() {}
This one is closed. It perfectly satisfies being sum types. It may not satisfy your opinion of what makes for good ergonomics, but if you want to talk about ergonomics let's use ergonomic words, not type theory words.
If C++ died for services under Unix and medium sized indie games (such as I2PD, or Cataclysm DDA and its forks) and these where rewritten in Go the maintenance would be far greater in these projects. A GC, a much better cross compatibility and your code for sure could probably be compiled in 10 years.
Also CDDA:Bn wouldn't damn need a > 40h long build using 1.5 GB of RAM under an n270 netbook. Ditto with Nchat, FFS, which is worse. A Golang counterpart for nchat as a TG client (and tdlib rewritten in Go) would weight far less while compiling and maybe even the binary itself, and performance wise it would be similar.
I still remember tons of C++ projects from 2005-2009 impossible to compile today but with GCC 4.3 or GCC 4.9, can't remember. Not because of the size, but because C++ incompatible changes over the years. At least tons of C code will compile it today as is modulo some POSIX changes from C code pre 1996. C++ it's something that should died long ago, among stuff like locales under Unix. UTF8 everywhere, use your own currency and the like.
Yeah, I know, game engines, and tons of physics engines and libraries even FLOSS ones such as OGRE, gdal and the like are C++ domain. Still, most of these could be ported to C.
The reality of such platform is observable in Inferno and Plan 9, that had no C++ compilers, and while Alef failed in Plan 9, Limbo came to be in Inferno.
Likewise, trying to write pure C or C++ services in iOS or Android is more pain than gain.
Two things: (1) I see that you are using a throwaway/new account. If throwaway, I have little sympathy for any downvotes that you get. If new, welcome to the community. I hope your share you personal experiences. (2) Nothing gets me more angry than telling highly skilled people it is a "problem between keyboard and chair." ("Oh, you just need to use it correctly.") As a top secret C++ fanboi for more than 25 years, I am just so tired of hearing this bullshit. As much as it hurts me to say it, Rust is better at FORCING programmers to do the right thing... instead of C++ where you CAN do the right thing. In my mind, without a very fast iteration for C++ "dialects" (see the Google project) where teams can trivially enable or disable language features (like multiple inheritance), C++ is a dying language compared to Rust.
there is a good chance Rust will start dying and will actually die by being replaced by some new hyper-overengineered lang much faster than C++ actually die.
The good news is as of now ladyboy doesn't have any competition.
Rarely if ever is anything able to compete simply by being "better". As far as USPs go it's just not enough. I reckon for ladyboy the USP (if any) is going to be it being open and NOT chrome (or derivative). So "safe" "modern" language is not going to mean much to the end users.
What's different today really is the LLMs and coding agents. The reason to never rewrite in another language is that it requires you to stop everything else for months or even years. Stopping for two weeks is a lot less likely to kill your project.
He's still right if you don't have good automated testing and you lost most of the original developers (or you don't have other seniors ceva familiar with the domain).
> then these two teams are essentially in a race against each other and the port will likely never catch up
Ladybird appears to have the discipline to have recognized this: “[Rust] is not becoming the main focus of the project. We will continue developing the engine in C++, and porting subsystems to Rust will be a sidetrack that runs for a long time.”
The context matters when we talk about Joel's article[0].
It's about Netscape. By the time, Netscape had dominated the browser market. It was the leader and that means they had all the market share to lose. You can bet Microsoft's decision makers were very closely monitoring what those at Netscape were doing.
Today, practically nobody uses Ladybird. No one even knows it[1]. It's so behind and has nothing to lose. If you really want to rewrite, it's better to do it when you have nothing to lose.
Firefox is already spying on you with a lot of telemetry, and they have recently amended their terms of use to remove the obligation to "never sell your data" [1]. So perhaps you should reconsider that statement.
Correct me if I’m wrong since I don’t know these two languages, but like some other languages, doing things the idiomatic way could be dramatically different. Is “cleanup” doing a lot of heavy lifting here? Could that also mean another complete rewrite from scratch?
A startup switching languages after years of development is usually a big red flag. “We are rewriting it in X” posts always preceded “We are shutting down”. I wish them luck though!