- C++11 is good enough (and a little better than that)
- The eco system. Yes, I can use C libs from dlang but managing extra bridges/headers is too much overhead if the alternative is just to throw them into my C++ project and use them without writing any glue code.
- The garbage collector. (RAII fan here)
- Tooling. Are there any good dlang IDEs with mature refactoring support and code completion?
For the tasks I'm using C++ for, C++11 is more than sufficient. For more abstract/high level stuff where I don't particularly care about top performance/memory layout I don't need another C++ contender but can use higher level languages/scripting languages.
D would look more tempting if there was no C++11 but for now for me I'm perfectly happy with C++.
- dub + VisualD makes building code, managing dependencies and debugging painless for me.
- being part of building the ecosystem is fun
- I'm not comfortable with the high cost of developing in C++ (even with last revision) and never chose to use it in the first place. C++ is basically forced on me since years.
- I feel like I can be more productive and write higher quality programs when using D. I can use RAII in D like I would in C++.
- I feel like Go and Rust do not address C++ shortcomings while introducing problems of their own.
- D programs are short. I don't need nor want IDE refactorings.
- compilation speed is not as impressive. To me this is imho the biggest pain point with C++, bigger then memory unsafety.
- I missed templates based on any value/symbol, it forced macros on me in my very first try. I don't think it's a problem once you know Rust, but having to write macros was too intimidating for a newb.
> compilation speed is not as impressive. To me this is imho the biggest pain point with C++, bigger then memory unsafety.
Compilation speed remains a work in progress; it's improved a lot in recent versions. Note that 80% of the compilation time in rustc is actually in LLVM passes (the same used by clang), which are mostly optimizations, so turning off optimization helps Rust compile times a lot. LLVM as a backend does many more optimizations than DMD (and 6g/8g for that matter) and runs slower as a result.
One issue is that Rust currently does not do incremental compilation below the crate level, so you will need to use many crates if you want incremental compilation. This will possibly change at some point via a ccache-like mechanism.
If memory safety is not important to your applications, though, then Rust may indeed not be the right choice. Rust requires you to make sacrifices (a learning curve and less flexible unique pointers) to ensure memory safety, and if you just want to write "new" and "delete" then Rust isn't going to be the best language for you. (That said, I think there is a strong argument to be made that for many classes of programs it is worth some pain to eliminate memory safety problems, because memory safety issues frequently lead to security vulnerabilities. For us as browser manufacturers, memory safety is so important it is worth sacrificing literally everything else except maybe performance in some cases.)
> - I missed templates based on any value/symbol, it forced macros on me in my very first try. I don't think it's a problem once you know Rust, but having to write macros was too intimidating for a newb.
In general you want to use traits instead of macros to write generic functions. There are some current limitations with traits but I think they're a more scalable approach than ad-hoc templates. It'd be helpful to know exactly what limitations you were hitting.
Ok I think I was unfair to Rust in my top message. "Creating problems of their own" was not accurate, but I am under the impression I would miss some D things if using Rust.
> Note that 80% of the compilation time in rustc is actually in LLVM passes (the same used by clang), which are mostly optimizations.
OK I had a flawed perception probably because of this.
> That said, I think there is a strong argument to be made that for many classes of programs it is worth some pain to eliminate memory safety problems
I agree 100%, especially on teams memory safety become incredibly important, also I have to thank you for fixing C integer promotion which I've seen create many bugs.
> It'd be helpful to know exactly what limitations you were hitting.
Ah, OK. From a glance I think some combination of higher-kinded type parameters, functional dependencies/associated types, and associated constants could fix the issues you were hitting that forced you to use macros. But I'll talk to bjz. Thanks for the pointers!
C++11 borrowed a few good ideas from D and other languages so that certainly helped close the gap between D and C++ but D is still significantly nicer to work in than C++ (which is what I spend all day at work writing).
You can use RAII all you want in D. The D standard library uses it for things like Files, containers, and smart pointers. I feel like Garbage Collectors have become something of a bogeyman for native languages. There is a lot of value a garbage collector can add. In D it's rarely hard to switch to managing your own memory if your profiler shows the garbage collector taking up more time than you'd like. For most programs avoiding the garbage collector is a premature optimization though.
What exactly is the argument for garbage collection over RAII-style freeing? Even in garbage-collected languages, you still need to do things like close file objects; the runtime doesn't do so automatically when it detects that there are no stack references to the file. Same goes for unlocking mutexes, or any of the other things that RAII tends to address.
Is it just that maintaining ownership semantics for heap memory tends to be more complicated than doing the same for file objects (much more rarely shared, I'd guess) or mutexes (not sure if "sharing" is even sensibly defined here).
There are a number of reasons why you may want garbage collection/automatic memory management over RAII:
- RAII-style freeing means a lot of reference counting. Reference counting and multiple threads don't mix well. You can circumvent that by having multiple separate heaps, but that's not always a practical solution.
- Reference counting is slow; if you want GC-like speed and RC-like pause time guarantees, you're probably better off with deferred reference counting, but that's not easy to engineer into C++; it's essentially a form of garbage collection.
- Lack of fully automatic memory management may affect modularity adversely; see Jones/Lins for the gory details. In brief, wholly or partly manual memory management creates leaky abstractions (this is if you eschew RC for some of your code/types for speed reasons or because you need to deal with cyclic references).
- It is not difficult to have scoped deallocation for resources (scope statement in D, using statement in C#, etc., destructor pragma in Nimrod). Resource lifetime, in any event, does not always coincide with the lifetime of a variable, so this is an incomplete solution. Using GC does not mean not using RAII where its downsides do not matter.
RAII does not mean a lot of reference counting. In the vast majority of cases in typical programs, objects have exactly one owner, and managing that ownership is very easy if your language provides good facilities for move semantics (as C++11 and Rust do).
You can't say reference counting is bad for multi-threaded programs without mentioning that GC is also bad for multi-threaded programs, since the garbage collector has to pause all other threads, either for the entirety of the GC run or (if you're lucky and using a good GC) for parts of it.
The modularity argument is reasonable, and I've been annoyed by this aspect several times, though in the majority of cases it doesn't seem to be an issue.
If you mix up reference counting with manual memory management, you get the aforementioned modularity issue (as I mentioned as one of my points). You get rid of (some) of the overhead, at the cost of simplicity (you also may incur other overhead; manual memory management often introduces unnecessary copying in order to keep ownership simple).
The problem with reference counting in concurrent programs is one of correctness, not one of speed [1] because every change of a reference count is a potential race condition. That's quite different from the challenges garbage collection faces in a multi-threaded environment.
Whether a garbage collector has to pause all threads is an implementation issue. Modern garbage collectors can limit that pause effectively to the root scanning phase, then let the collector run concurrently with the mutator. You can also work around pausing threads entirely, though the tradeoffs are rarely worth it outside of hard realtime environments.
Note also that this is an issue of pause time, not performance. While HN can get obsessed with pause times, not everyone programs video games, embedded systems, or kernel drivers where that matters. High-performance computing, for example, is a huge field where pause times are all but irrelevant and where amortized cost is what matters. (HPC is also where correctness can become easily more important than squeezing out a percent or more of performance through low-level hacks; if a job that takes several days to run crashes halfway through, that's an enormous loss.)
[1] Technically, you can make reference count updates atomic, but the overhead is absurd, especially under contention. Hence why SNZIs [2] exist, which reduce the speed overhead at the expense of additional memory overhead.
[2] http://dl.acm.org/citation.cfm?id=1281106
>While HN can get obsessed with pause times, not everyone programs video games, embedded systems, or kernel drivers where that matters.
The (my) problem is that it also matters whenever a program holds lots of data in memory (databases, data analysis, caching, etc). As the cost of RAM decreases, the importance of this problem increases, i.e. fast.
Large heaps and large pause times do not necessarily go hand in hand. That's a question of GC technology. For example, the Azul GC does <10 ms pauses for heaps that are hundreds of GB in size. Granted, Azul/Zing pricing doesn't exactly make it the most accessible technology (and it relies on kernel hackery for its compaction), but it demonstrates that multi-second pauses are hardly a necessity.
Incremental garbage collection isn't a new technology [1] or one that's particularly difficult to implement by itself (a grad student could probably implement Baker's treadmill in a couple of days); what makes it hard is primarily compaction and multiple threads [1]. You can't effectively use (naive) RC with multiple threads and you don't get compaction with RC, either. Multi-threading is in practice a major driver for the need of GC, since (naive) RC isn't thread-safe and unpredictable object lifetimes don't go well with manual memory management.
Also, deferred RC strategies can under certain circumstances contain the cost for cycle collection. Trial deletion is already limited to nodes reachable from other nodes whose reference count has been decremented; type information can be leveraged further to exclude nodes that cannot possibly be parts of cycles (this is particularly easy in ML-style languages, which make mutually recursive types explicit in the source code, but is not limited to those [2]).
Finally, you can also use multiple disjoint heaps to simplify implementation and cap GC cost (one per thread and zero or more shared heaps). This can also bring naive RC back into play as a viable strategy, though you'd still lose compaction. Multiple heaps are particularly attractive for NUMA architectures.
[1] I note that hard realtime guarantees are difficult to make with a basic incremental collector due to the potential of pathological behavior, but we are not talking about hard realtime guarantees here.
[2] Obviously, JITs and dynamically or weakly typed languages have a harder time with this approach.
>Large heaps and large pause times do not necessarily go hand in hand
Not necessarily, but in practice they do go hand in hand. I didn't say the problem was impossible to solve, just that it is an important problem that needs solving. Azul solved the technology issues (or so they claim) but not the economics of it, and they solved it for a language that isn't a good fit for in-memory computing in the first place (to put it politely).
If I have to write software today that keeps a lot of data in-memory and requires reasonable latency (and I do) my only realistic option is C++.
I know all the drawbacks of naive reference counting. C++ shared pointers are a horrible kludge. Fortunately they are only needed in very few places, thanks to RAII and unique_ptr. The problem is that C++ has tons of other issues that will never be solved (antiquated modularity, header files, crazy compile times, excessive complexity and generally lots of baggage from the past).
If I have to write software today that keeps a lot of data in-memory and requires reasonable latency (and I do) my only realistic option is C++.
I don't necessarily have a fundamental disagreement here, but I offer two caveats (one of which you may consider disagreement at a certain level).
One is that a lot of the discussion in this thread is not about what is possible today, but about where language implementations can realistically go.
The second is that there are language implementations that do allow you to keep lots of data in memory and still give you low latency; the catch is that most of them are functional programming languages and do not tie into the ecosystems of C++, Java, etc. which limits their applicability. But consider for example, that Jane Street Capital, which has pretty significant requirements for latency, is using OCaml. This is in part because OCaml has an incremental garbage collector for major collections (in addition to generational garbage collection). As I said, it's not rocket science.
The same goes for Erlang, which uses a model of thread-local heaps. Erlang heaps tend to be small (lots of threads [1] with little data each, even though the total memory footprint may be in the gigabytes), so Erlang can use a straightforward garbage collector that can run concurrently with other threads, does a complete collection fast (because the entire heap fits inside the L3 or even L2 cache) or can avoid collection entirely (if you specify the heap size explicitly with spawn_opt and no collection is needed before thread termination). As a result, Erlang can easily satisfy the low latency requirements for telecommunications (where its being used primarily).
Functional programming languages just had an unavoidable need for garbage collection for half a century now and so implementations of functional languages have seen a lot more GC tuning than imperative languages. Thus, you do at least in theory have other options, but I expect "realistic" in your case also implies being able to tap into certain existing software ecosystems.
Let me finally note that a big problem has arguably been too much focus on the JVM; not that there's anything wrong with supporting the JVM, but it has too often come at the cost of alternative execution models. JIT compilation, lack of value types, a very general threading model, loss of static type information, etc. all can very much get in the way of tuning automatic memory management. Luckily, some of the newer emergent languages target alternative backends, in part to avoid these problems.
[1] Technically, Erlang uses the term "process" in lieu of "thread"; I'm using "thread" to avoid confusion with OS-level processes.
Modern GCs use multiple threads to collect garbage. Also, it is possible to design a GC that does not pause mutator (e.g. C4) or does most of the things concurrently (CMS) or does incremental sweeps (G1 and other evacuating collectors), or uses separate isolated heaps for each thread (Erlang GC).
There are some non-defered reference counting that can offer pretty good performance(within 19% of defered reference counting), while still offering determinism[1].
Yes, but they have the same problem that they require compiler smarts; i.e., they require more than a template to implement shared_ptr<T>. When you're doing that, you may just as well implement something more sophisticated for automated memory management.
The attraction of RAII is that it's a pretty simple mechanism that doesn't require compiler intervention outside of basic, already existing optimizations such as inlining.
Lack of GC makes concurrent lockless programming extremely hard (and most of the examples on the internets are subtly broken). RAII doesn't work at all besides trivial cases when there is no sharing at all. Reference counting incurs additional contention, due to interlocked increments/decrements and is also very hard to do properly. Where in a GCed language a single CAS is often enough, in a non-GCed language you need to resort to less efficient and often completely unsupported things like DCAS or CAS2.
There have been papers on how GCs can be more efficient than manual memory management in some situations, so it's not always the case that RAII is better from a performance standpoint (though I think that it frequently is). But regardless, a big reason for the GC in D is memory safety. Using the GC, you can make memory safety guarantees that cannot be made with malloc and free (e.g. you can guarantee that a pointer won't be invalid with a GC but can't with manual memory management), and D is big on making memory safety guarantees. Also, some features (such as array slices or closures) work very well when you have a GC and become much more difficult to pull off without one due to a lack of automatic memory management that the compiler can use and a lack of clear ownership in some cases (e.g. the runtime owns the memory for array slices rather than there being a particular array which owns the memory).
That being said, D does not require the GC. It's easier if you use it for at least some of the features, but D makes it very easy to use RAII and manual memory management, which also helps a lot in making the GC work better, because it's much easier to avoid making a lot of garbage for it to collect when you don't need to. A lot of stuff in D ends up on the stack rather than the heap, and D's GC ends up having a lot less work to do than the GC does in languages like C# or Java. That being said, the current GC implementation needs some work (and effort is definitely put forth in that area), but if you don't want to use the GC, you can minimize its use or even outright avoid it (though _completely_ avoiding it can be a bit annoying since that means avoiding a few language features that require it; that list is short though).
So, while D definitely uses the GC and promotes its use where appropriate, you have full control over memory just like you would in C++. And the few features that are hampered by avoiding the GC don't even exist in C++ in the first place.
> A lot of stuff in D ends up on the stack rather than the heap, and D's GC ends up having a lot less work to do than the GC does in languages like C# or Java
This is true, however there's a compromise involved. Because D also allows for manual memory management and unsafe memory access, it means that the GC is not free to move stuff in memory at will ... which really means that garbage collectors, like the ones available for Java (precise, generational, non-blocking, fairly predictable and compacting) are very hard to develop, probably next to impossible. This is the price you pay for using a lower level language and it's not a price that will go away easily.
I've been using Scala a lot for the past two years, using it to build high-traffic web services and while Scala is pretty wasteful in terms of allocating short-lived objects, I've been surprised at how well the JVM handles it.
In terms of throughput for example, Java GCs are much better than manual memory management. Allocating memory usually involves just incrementing a pointer, so it's basically as cheap as stack allocation. Deallocating short-lived objects is also fairly cheap, since it happens in bulk and so the amortized cost is pretty similar to dealocating stuff on the stack (!!!). The JVM can do some pretty neat things, like for example if it detects that certain references do not escape their local context, it can decide to allocate those objects straight on the stack.
What really sucks about garbage collection is the unpredictability. Java's CMS for example, awesome as it is, still blocks the world from time to time. And when it does, you have no real control over how much time it keeps the process hostage. The new G1 in JDK7 is much better and if you want the state of the art in terms of near-real-time GCs, you can buy into Azul's pauseless GC. But they still suck for certain apps. Allocating objects on the heap also means you have to pay the price of boxing/unboxing the references involved. This sucks too.
On the other hand, by having a good GC at disposal, it's much easier to build multi-threaded architectures. In C++ for example, it's so freaking painful to deal with non-blocking concurrent algorithms, or really, multi-threading of any kind.
> Because D also allows for manual memory management and unsafe memory access, it means that the GC is not free to move stuff in memory at will
Since I wrote a moving GC for Java years ago, and hence know how they work, I set the D semantics so it allows a moving GC. It's just that nobody has written one for D.
It is hard to manage ownership when allocations are passed around between concurrent threads. Garbage collection in Go makes goroutines and channels easy.
Though arguably you shouldn't be accessing the same allocation from concurrent threads at the same time anyways, because that requires mutexes, which are ugly. Rust requires you to transfer ownership of objects when passing them between threads, which both enforces mutex-free programming and makes RAII work. Only one thread owns an object at a time, and that's the only thread that can use the object, and is also the thread that is responsible for freeing it when it goes out of scope.
Mutexes are only ugly because few languages properly support them. It is not difficult to ensure at the language level that no object is accessed without owning the the lock associated with that object (read/write locks can be a bit trickier), but for some reason hardly any language implements something along these lines.
It's not fundamentally different from having bound checks for array indices. You can either do that pretty easily at runtime or use a simple type system to do it at compile time (e.g., Boyapati's [1] or Flanagan's [2] work). It's been done, e.g. for Cyclone [3]. This is decades old stuff, not rocket science. Honestly, Monitors had the basic idea right in the 1970s, except as Per Brinch Hansen lamented [4], language designers keep screwing it up.
This implies that an object will be unavailable while it is being passsed between threads. How fast is that? With a mutex, it's as fast as shared memory.
Message passing is a couple of atomic operations, and there are also L4-esque optimizations that allow a message send to be optimized into a direct context switch without a pass through the scheduler.
Also, you can use shared memory in Rust. If you do, the type system ensures there can be no data races: either the data is immutable or you must take the lock before you mutate the data.
Under the hood, you still use shared memory in this style of programming. But the programmer never uses shared memory directly. Instead he passes references to the shared memory to other threads, and the act of passing a reference clears the reference in the sending thread so the sending thread can no longer access it.
In Rust, this style is enforced by the language. In C++11 you can enforce it by the convention of using std::unique_ptr and passing those between threads via a shared data structure like a blocking queue.
> Even in garbage-collected languages, you still need to do things like close file objects; the runtime doesn't do so automatically when it detects that there are no stack references to the file.
I'd just like to say there are sane solutions out there already to things like this. With the proper abstractions you get things like custodians[0] in Racket and several alternatives in terms of "call-with-..."[1] functions that do handling for you.
These might not be the languages that are being discussed here, but they offer things like these precisely because they are not the languages discussed here right now.
For reference, a lot of embedded programmers still use static allocations only, not even trusting heap allocations at all.
Took awhile to get used to, but it is a damned fine way to avoid memory leaks of any kind!
But even that aside, if you are writing something like a render loop with hard timing guarantees, (or anything else with timing needs!) having non-deterministic allocation costs can easily blow your cycle budget.
"defer" has really weird semantics; it's based on imperatively pushing cleanups onto implicit per-function mutable state. (This shows up if you call "defer" in a loop, for example.) It's also possible to screw up and forget to use "defer", causing things like leaks of file handles. RAII, on the other hand, is based on lexical scope, without hidden mutation, and also makes it harder to forget to do things like close files.
It's more complex to specify and implement than "this code statically runs at the end of the block". This can lead to surprises. (I would not have guessed what "defer" in a loop does until I learned.)
Because "defer" is tied to function definitions, it means that pulling code out into a new function or inlining a function will silently change its behavior in subtle ways.
Like most complex language semantics, "defer" is less easily optimizable than the simple semantics of RAII—with RAII the compiler statically knows what it is supposed to call, which helps exceptions (panicking goroutines in Go) because it allows their cleanups to be optimized into table-driven unwinding. But because Go chose "defer", the compiler must insert code that dynamically keeps track of cleanups to handle the general case. This results in indirect function calls, and bookkeeping overhead, whereas RAII is purely static function calls. The bookkeeping overhead is significant: it allows RAII to be zero-cost in the no-panic path (very important in C++'s case!), while I don't see any way to implement "defer" in a zero-cost way.
defer is great for "one off" cleanups, but in so many cases (closing file handles, releasing locks, etc.) cleanup is not one off but is repeated over and over again. So the price of explicitness is redundant code that is easy to forget or get wrong.
Note that if your language supports RAII and lambdas (e.g. C++11), it's trivial to implement defer using a class that takes a function as an argument to its constructor and calls that function in its destructor. So you can still have the flexibility of defer in a language with RAII.
My preferred method of handling this is how Ruby does it with, e.g., File.open. The function yields the file handle and then closes it once the supplied block (think lambda function) ends. This is similar to what you refer to, but without the class. The class is useful for heap objects, of course.
- C++11 is good enough (and a little better than that)
- The eco system. Yes, I can use C libs from dlang but managing extra bridges/headers is too much overhead if the alternative is just to throw them into my C++ project and use them without writing any glue code.
- The garbage collector. (RAII fan here)
- Tooling. Are there any good dlang IDEs with mature refactoring support and code completion?
For the tasks I'm using C++ for, C++11 is more than sufficient. For more abstract/high level stuff where I don't particularly care about top performance/memory layout I don't need another C++ contender but can use higher level languages/scripting languages.
D would look more tempting if there was no C++11 but for now for me I'm perfectly happy with C++.