I have no constructor, and I must initialize(consteval.ca) |
I have no constructor, and I must initialize(consteval.ca) |
T̶h̶a̶t̶ d̶o̶e̶s̶n̶'t̶ s̶e̶e̶m̶ c̶o̶r̶r̶e̶c̶t̶:̶ a̶ d̶e̶f̶a̶u̶l̶t̶e̶d̶ c̶o̶n̶s̶t̶r̶u̶c̶t̶o̶r̶ s̶t̶i̶l̶l̶ d̶e̶f̶a̶u̶l̶t̶-̶i̶n̶i̶t̶i̶a̶l̶i̶z̶e̶s̶ t̶h̶e̶ m̶e̶m̶b̶e̶r̶s̶, n̶o̶t̶ v̶a̶l̶u̶e̶ i̶n̶i̶t̶i̶a̶l̶i̶z̶e̶. I̶ d̶o̶n̶'t̶ t̶h̶i̶n̶k̶ t̶h̶e̶r̶e̶ i̶s̶ a̶n̶y̶ d̶i̶f̶f̶e̶r̶e̶n̶c̶e̶ b̶e̶t̶w̶e̶e̶n̶ d̶e̶f̶a̶u̶l̶t̶i̶n̶g̶ i̶n̶l̶i̶n̶e̶ a̶n̶d̶ o̶u̶t̶ o̶f̶ l̶i̶n̶e̶. G̶C̶C̶ s̶e̶e̶m̶s̶ t̶o̶ a̶g̶r̶e̶e̶:̶ h̶t̶t̶p̶s̶:̶//g̶c̶c̶.g̶o̶d̶b̶o̶l̶t̶.o̶r̶g̶/z̶/r̶4̶r̶e̶5̶T̶E̶5̶a̶
edit: I missed that the author is actually value-initializing x!!! The result definitely violates expectations!
Generally, the details of the rules are arcane and sometimes have non-sensical dark corners having been extended and patched up for the last 40 years. But 99.9%[1] of the time you get what you expect.
I big improvement would be making default initialization explicit, and otherwise always value initialize. Explicit value initialization is so common that the very rare times I want default initialization (to avoid expensively zeroing large arrays) I need to write a fat comment. Writing "std::array<int, 100> = void;" (or whatever the syntax would be) would be much better.
[1] I had an extra 9 here... I hedged.
Actually initializing your instances, which is what's expected of every single instantiation, is enough to not experience any problem. This also means that if you want to call a constructor, you need to define it.
This is a case of people trying to be too clever for their own sake, and complaining that that's too much cleverness for them to handle.
Edit: this response is a bit dismissive but honestly my main beef with this article is that its conclusion is just straight up wrong. Do not write your own constructors, do follow the rule of 5/3/0, and if you find yourself needing to hold a const reference, you should look out for whether you’re passing in an rval temporary… none of this is really scary.
> In my humble opinion, here’s the key takeaway: just write your own fucking constructors! You see all that nonsense? Almost completely avoidable if you had just written your own fucking constructors. Don’t let the compiler figure it out for you. You’re the one in control here. Or is it that you think you’re being cute? You just added six instances of undefined behaviour to your company’s codebase, and now twenty Russian hackers are fighting to pwn your app first. Are you stupid? What’s the matter with you? What were you thinking? God.
The problem with C++ and the danger with an article like this is someone might actually follow this advice, instead of eg: the core guidelines.
Every other example is a violation of the core guidelines in some form or another. There is no other problem.
That's my take as well. The blogger clearly went way out of his way to find something to whine about while purposely ignoring a few of the most basic principles and guidelines.
In the meantime, everyone who ever went through a basic tutorial just casually steps over these artificial scenarios.
Your statement may be correct! But it’s certainly not common knowledge in my experience.
https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines...
https://isocpp.github.io/CppCoreGuidelines/CppCoreGuidelines...
std::reference_wrapper still can’t save you from yourself, but it’s better than violating the first link and ending up in this limbo that OP is talking about.
See also: https://youtu.be/YxSg_Gzm-VQ (about 3:30 in)
Nothing else.
https://talesofmytery.blogspot.com/2018/10/harlan-ellison-i-...
> So, here’s the glue between list-initialization and aggregate initialization: if list-initialization is performed on an aggregate, aggregate initialization is performed unless the list has only one argument, of type T or of type derived from T, in which case it performs direct-initialization (or copy-initialization).
The word “unless” is even bold.
We have fancy syntax:
T t{v0};
T t{v0, v1};
So you can write your own constructors, and you can initialize a tuple or array with only one element supplied, and you might trip over the wrong constructor being invoked in special cases.
I remember discovering this when C++11 initializer lists were brand new and thinking it was nuts.
It's 15 years out of date now, but also timeless since C++ rarely/never removes old features or behaviours.
I'm hoping something like Herb's C++ syntax 2 will make the language useable for mortals like me.
I think you're whining about something that doesn't pose any problem to anyone with any passing experience in software development.
The examples in the blog post boil down to far-fetched cases devised to trigger corner-cases of a feature where a programming language in exceptional cases auto-generates specific functions when programmers somehow explicitly decided not to do it themselves. The blogger then proceeds to explore edge conditions that lead these exceptional cases to either be enabled or disabled.
In the meantime, be mindful of the fact that this is about a language with a clear design goal of paying only for what you use,and also the widely established rule of 3/rule of 5, which is C++101 and states that when anyone defines one of these special member functions, they should define them all. Why? Because it's C++101 that these special member functions are only generated by the compiler in specific corner cases, and given their exceptional nature the compiler will not generate them automatically if any of the requirements is not met.
Therefore, any programmer who goes through the faintest introduction knows that they have to set the constructors they will use. Is this outlandish?
Also, does it catches anyone by surprise that you need to initialize instances when you instantiate them? Is this too much of a gotcha to justify being called "horrific"?
I think people like you just feel the need to have something to complain about. In the meantime, everyone in the real world happily does real work with them without any fuss.
Thanks for the ad hominem
It’s been a while now, and at least in my experience so far Go and Rusts choice of not having special constructors really simplifies a lot.
Is there anyone that’s had the experience of missing constructors once you swapped away from them?
Such as all the constructors that are being added, implicit copy constructor and all the other surprises?
T::T() = default;
Link: https://consteval.ca/2024/07/03/initialization/#:~:text=You%...
That actually isn't that weird, because it would allow any consumer of your library to change how your library behaves.
Kudos to that author for the great, eye catching title and the in depth detail!
That can't be right ... is it? Things cannot be initialized twice. Isn't it more like "Otherwise, recurse the value-initialization over the bases and members". Then, those that are not classes or arrays get zero-initialized.
It doesn't mean that the lifetime of the object starts twice, or anything weird like that.
You can only initialize once. After it's been initialized you're just assigning values, and that's not what happens during initialization. It's either a misunderstanding on behalf of the author or the words as written are not conveying the correct idea.For good or for ill, I don't really trust anything that long to be something the average human can wrestle down. And when you mix in undefined behavior and the fact it's still used on safety-critical systems... It's a bit horrifying really.
I wonder how much shorter the spec for a radically simplified version of C++ -- say, one that zero-initializes everything, and has no default constructors ("not writing a constructor raises an error" or something) -- would be? 5% 50? 95...?
So their claim that "T t;" will "do nothing" is incorrect.
class T
{
public:
T(int);
};
T t;
> Primarily, there are two kinds of initialization of concern: default-initialization and value-initialization. The rules given in the standard look roughly like this:
> * For any type T, T t; performs default-initialization on t as follows: ...
As GP mentions, the article's descriptions of default and value initialization are both incorrect for classes that do not have default constructors, as that code will simply not compile.
I like keeping the rules of the language simple enough that there is never any confusion.
"Call J. G. Wentworth, 877 Cash Now!"
Unless you want to `ptr::write` individual fields by hand into a `MaybeUninit`, which you can absolutely do mind but that… is not very ergonomic, and requires structs to be specifically opted into this.
Without guaranteed “placement new” that can mean that your 2MB object gets constructed on the stack and copied to the heap. And while Linux defaults to a 4MB stack, Windows defaults to 1MB and will crash your program. Or it might work if the compiler optimizes in your favor.
It's not something you encounter frequently, it can be worked around, and Rust will eventually solve it ergonomically without introducing constructor hell (probably with just a keyword). But finding the best language-level solution isn't straightforward (efforts to fix this for rust are ongoing for 9 years)
This take makes no sense. Think about it: you're saying that not having the compiler do any work for you "really simplifies things a lot". Cool, so you have to explicitly declare and define all constructors. That's ok. But think about it, doesn't C++ already offer you that option from the very start? I mean, you are talking about a feature in C++ that is not mandatory or required, and was added just to prevent those programmers who really really wanted to avoid writing boilerplate code to lean on the compiler in and only in very specific corner cases. If for any reason you want the compiler to do that work for you, you need to be mindful of the specific conditions where you can omit your own member functions. For the rest of the world, they can simply live a normal life and just add them.
How is this complicated?
Complaining that special member functions make obvious things less simple is like complaining that English is not simple jus because you can find complicated words in a dictionary. Yes, you can make it complicated if that's what you want, but there is nothing forcing you to overcomplicate things, is there?
All structs in Rust must be initialized using brace syntax, e.g. `Foo { bar: 1, baz: "" }`. This is commonly encapsulated into static functions (e.g. `Foo::new(1, "")`) that act similarly to constructors, but which are not special in any way compared to other functions. This avoids a lot of the strangeness in C++ that arises from constructors being "special" (can't be named, don't have a return type, use initializer list syntax which is not used anywhere else).
This combined with mandatory move semantics means you also don't have to worry about copy constructors or copy-assignment operators (you opt into copy semantics by deriving from Clone and explicitly calling `.clone()` to create a copy, or deriving from Copy for implicit copy-on-assign) or move constructors and move-assignment operators (all non-Copy assignments are moves by default).
It's actually rather refreshing, and I find myself writing a lot of my C++ code in imitation of the Rust style.
Here are some ways:
- As a junior programmer, it made the language harder to learn. Language complexity increases super-linearly as each new feature has rules of interaction with several existing features
- Although one eventually learns to avoid the anti-features, you cannot control the actions of others. These features meant to help save keystrokes are happily employed every day, producing hard to read code
Particularly when writing library code for other to use or when maintaining large codebases shared by hundreds of engineers, my experience is that complex features in the language end up used by junior engineers or require consideration in API design.
and from what I understand rust constructors are basically the same as java, no?
Rust has functions associated with types which are conventionally used like constructors, but critically the new objects must have all their fields provided all at once, so it is impossible to observe a partially initialised object.
[0] https://learn.microsoft.com/en-us/dotnet/fundamentals/code-a...
[0] except in the more generalised haskell-ish sense that structs or enum variants can be constructed and some forms (“tuple structs” and “tuple variants”) will expose an actual function
Eg:
- Java constructors can return the object before they complete construction, finishing at a later time; this is visible in concurrent code as partially constructed objects
- Java constructors can throw exceptions and return the partially constructed object at the same time, giving you references to broken invalid objects
- Just.. all the things about how calling super constructors and instance methods interleaved with field initialization works and the bazillion ordering rules around that
- Finalizers in general and finalizers on partially constructed objects specifically
I don't in any way claim it's on the same level as C++, but any time I see a Java constructor doing any method calls anymore - whether to instance methods or to super constructors - I know there are dragons
https://codefibershq.com/blog/golang-why-nil-is-not-always-n...
So horrifyingly true.
1: https://en.wikipedia.org/wiki/I_Have_No_Mouth,_and_I_Must_Sc...
You have "C with classes" that coexist with the "modern" way, full of smart pointers and functional programming. It is popular in embedded systems, video games, servers, and GUIs (mostly Qt). And if you look at the code, it is as if it was a different language, because the requirements are all very different. Embedded system need low level hardware access, video games are all about performance, servers want safety, and GUIs want flexibility.
There are less awful alternative to C++. For example C on one end of the spectrum and Rust on the other end. But none of them cover every C++ use case.
- There’s at least an ugly library to do it in C++
- There might be support baked directly into the language
- Or you could do it in Lisp, but that would be too easy
Well, it does unstructured imperative, structured imperative, and OOP imperative!
Except if you count template programming, because that one is pure functional, but only runs at compile time.
Literal lol... this is not an argument in favor of C++.
The fact that you can do almost anything IS pretty cool, but without having at least one C++ wizard at hand it can drive you nuts.
If I was starting a new work project with a lot of junior team members, or if I was doing a web project, or a very simple script, fine I’ll use a different language. There can definitely be good reasons not to use C++. But I’m at the point in my expertise that I will default to C++ otherwise. I’m most productive where I am most familiar.
Not disagreeing that C++ is awful in a lot of ways and super difficult though. But I still weirdly like it, personally. I find it a fun challenge/puzzle to work with.
In any case, I think the primary goal of any programming language is to get out of your way and let you tackle more interesting problems related to the problem domain that led you to start writing a program in the first place.
You're basically saying the language gets in the way of solving your problem :)
No other language creates as many problems for programmers as C++.
but only if not knowing how the complicated parts work doesn't create any subtle issues and has reasonable compiler time errors and isn't fundamental needed to write any code
Because then you can use the language without needing to know how exactly that complexity works and if you get it wrong you get a reasonable compiler error. And then decide to either spend some time to learn what you didn't know or you just write the code differently. But in either case you don't have a situation with a unexpected runtime error which should be impossible and where you have no good way to know where to even start looking.
That being said it’s way less complex than C++’s rules and that’s welcomef.
However I can't help but think that maybe I'm just so fortunate to be able to work in a nice code base optimized for developer productivity like this. C++ is really a nice language for experts.
to enforce all fields initialized.
If you care, the linter is there, so this is more of a skill issue.
The Default typeclass (or trait) as seen in Haskell and Rust, is a far better design, as this makes the feature opt-in for data types that truly support them.
First example: https://godbolt.org/z/898rorEqG
Second example: https://godbolt.org/z/K6aceesG9
Third example: https://godbolt.org/z/KcbqzMzdK
and I can't get the fourth example to compile without fixing the problems.
CW though, it is pretty grim. Very early example of the "AI takes over the world, decides humans are redundant" trope though. (Maybe the first?)
Actually, strike that: I'm not sure if it's true or not (though I suspect it is), but it doesn't actually matter. What I'm really getting at here is that there is nothing in Rust that behaves so confusingly or ambiguously as what's described in this article. If you're writing Rust, you'll never have to remember these sorts of rules and how they are going to be applied to your code.
I do agree that reading someone else's Rust can be a challenge, if they're using Rust's type system to its fullest, and you're (quite reasonably and understandably) not up to speed on the entirety of it. And that is a problem, agreed; though, at least, fortunately it's not a problem of ambiguity, but more of a learning-curve issue.
But I have never been writing Rust and have been confused about what the code I'm writing might do, never had to consult and puzzle out some obscure documentation in order to ensure that the code I was writing was going to do what I expected it to do. C++ falls so incredibly flat in this department, and that's why I avoid using it like the plague.
https://github.com/titzer/virgil/blob/master/doc/tutorial/Cl...
There are 3 which pertain to object initialization in Java.
1. super is initialized in it's entirety by an implicit or explicit call to `super()`
2. All instance initializers of the present class are invoked in textual order.
3. Constructor code following the `super()` call is executed.
The only awkward thing here is the position of #2 in between #1 and #3, whereas the text of a constructor body suggests that #1 and #3 are consecutive. It gets easier to remember when you recognize that, actually, there's a defect in the design of the Java syntax here. A constructor looks like a normal function whose first action must be a `super()` call. It's not. The `super()` call is it's own thing and shouldn't rightly live in the body of the constructor at all.
Edit: Tweaks for clarity.
Any language that has constructors has some complex rules to solve those things. And it's always good to check what they are when learning the language. Java has one of the simplest set of those rules that I know about.
Java constructors do not actually return the object. In Java code, it would appear to the caller as though the contructor returns the new instance, but that is not really the case. Instead, the new object is allocated and then the constructor is called on the object in (almost) the same manner as an instance method.
Additionally, Java constructors can only leak a partially initialized object if they store a `this` reference somewhere on the heap (for example, by spawning a thread with a reference to `this`). The assertion that this gives you a reference to a "broken invalid object" is only potentially correct from the perspective of invariants assumed by user-written code. It is perfectly valid and well-defined to the JVM.
> - Just.. all the things about how calling super constructors and instance methods interleaved with field initialization works and the bazillion ordering rules around that
This is a gross mischaracterization of the complexity. There is only a single rule that really matters, and that is "no references to `this` before a super constructor is called". Until very recently, there was also "no statements before a super constructor is called".
> - Finalizers in general and finalizers on partially constructed objects specifically
Finalizers are deprecated.
> Java constructors can throw exceptions and return the partially constructed object at the same time
Can you elaborate on your opinion? I mean, I don't think that argument makes any sense. You're talking about an optional feature that, under very specific circumstances, you can get the compiler to fill in for you default implementations for factory methods/constructors.
As a junior developer, it should be very clear to you that if you want to call a function, including copy constructor or copy assignment operators, you need to define them first. Is that too much of a burden to place on a junior developer?
> Although one eventually learns to avoid the anti-features (...)
There are none of those, and obviously special member functions don't pose a problem to anyone.
> Particularly when writing library code for other to use or when maintaining large codebases shared by hundreds of engineers, my experience is that complex features in the language end up used by junior engineers or require consideration in API design.
I don't think you have a good grasp on the subject. I've worked on C++ libraries shared by hundreds of engineers, and API design was always from the start the primary concern. This is not a function of seniority: it's the very basics of writing modular code intended to be consumed by third parties.
Still, special member functions are the very least of anyone's concerns because anyone remotely competent in this domain knows very well that the public interface needs to be explicitly designed to support or reject specific types of uses, and the decision of whether a component could/should be copied/moved is tied to the component's architecture and semantics.
I never thought that constructors were that burdensome and therefore do not understand the omission in other languages like Go and Rust that followed. Quite the opposite really -- knowing that a type always went through a predefined init was comforting to me when writing Java.
If you want to simplyfing by synthesising decades of accumulated knowledge into a coherent language, or to remove depreciated ideas (instead of the evolved spaghetti you get by decades of updating a language) then fine. If your approach to simplicity is to just not include the complexity, you will soon disciplinary that the complexity was there for a reason.
I mean, a standard (and stupid IMO) interview question is rate your C++ expertise from 1-10, and if you answer more than about 5-6 you get bounced for lying or not recognizing your limitations, while they gleefully point out Stroustrup wouldn't answer 9-10.
Bolted on, very terrible OO that should never be touched
Some of the most batshit insane ideas of encapsulation anyone has ever seen
Some of the most batshit insane return rules anyone has ever seen
Encouraged inconsistency from the language in the form of functions like “Len” that are added because sometimes someone feels it reads better?
Encouraged unwinding exceptions as regular flow control (lol. Yikes)
It is nearly universally agreed that Python has significant code management issues as code bases scale
This is all ignoring debates of fundamental typing issues. Personally, I hate what Python does, but some people seem to prefer it.
Let us not pretend Python doesn’t have some language problems on top of its tooling problems.
That's the worst part. But you have to agree it's the best for throwing small sized glue code around.
> Sure, Python is slow, packaging/versioning is painful
Those are complaints. That is evidence that people complain about Python. You just did it yourself.
But maybe your complaints about C++ are an order of magnitude more plentiful than for Python. And maybe quite a few of your C++ complaints are about much worse things. But that's not the point: they are all complaints.
And that's the problem with the Stroustrup quote: he's implicitly saying that all complaints are created equal, and there's no difference between having 10 complaints or 10,000 complaints (where 3 of the first are major, and 5,000 of the second are major).
It's used, as the GP points out, to shut down legitimate complaints. "Oh, you don't like $REALLY_BIG_HORRIBLE_ISSUE with my language? Psh, whatever, people complain about all languages, I dare you to find another language that you won't find something to complain about." Not the point! Is $REALLY_BIG_HORRIBLE_ISSUE a problem or not? If not, actually explain and justify, with specific arguments, why you don't think it's a problem. And if you do agree it's a problem, stop deflecting attention, admit that's it's a problem, and try to find a solution!
- If T is a class type and there is a default constructor, run it.
- If T is an array type, default-initialize each element.
- Otherwise, do nothing.
That decision tree should read: "If T is a class type: it will invoke the default constructor. It is a compile-time error to write this if T does not have a default constructor." Not "if there is a default constructor, run it; otherwise, fall back to doing nothing."
The "do nothing" applies to scalar types such as ints, and indirectly to scalar member variables that aren't explicitly initialized one way or another. Not to classes that have deleted the default constructor.
... Which is not surprising, given that we're talking about a language with a spec that dwarfs most works of human literature for sheer mass.
I don't think you managed to understand what I actually said, and consequently you wrote a whole wall of text that's not related to the point I made.
I did not. Read what I wrote.
``` //explicitly annotating this struct is default initializable and copyable #[derive(Default, Copy, Clone)] struct Foo { ... } ```
is actually worse than
``` struct Foo {...}; // rule of zero, copy/move/default are defined/deleted based arcane rules predicated on the contents of Foo ```
If you read what I wrote you'll notice I was pointing out the absurdity of claiming that being forced to write each and every single factory method/constructor is somehow better and simpler than allowing the compiler to write them for us for trivia classes but still having the compiler step off when we opt to write each and every single factory method/constructor ourselves.
Compare `int albumId, songId;` versus `AlbumId albumId; SongId songId;`. The former two variables can be assigned to each other causing potential bug and confusion. The latter two will not. Once you have a basic wrapper for integers, further wrappers are just a one-liner so why not. And also in practice making the type more meaningful leads you to shorter variable names because the information is already expressed in types.
Most compilers will do something like "treat the bits like they represent a float, even though they mean something else when they're treated as an int."
But the language spec says the compiler is allowed to send an email to Bjarne Stroustrup with your home address so he can come over and personally set your computer on fire.
Here's an article on the topic: https://cryptoservices.github.io/fde/2018/11/30/undefined-be...
>Without guaranteed “placement new” that can mean that your 2MB object gets constructed on the stack and copied to the heap. And while Linux defaults to a 4MB stack, Windows defaults to 1MB and will crash your program. Or it might work if the compiler optimizes in your favor.
C gets a lot of hate, often for good reasons, but at least you know where your memory is coming from when you are allocating it yourself. If you're allocating a large heap-allocated object, you're grabbing the memory directly from the heap.
Those are temporary pains that have remedies in the works. Rust is a fairly young language, and a lot of good-enough solutions get thrown out before ever getting beyond the experimental stage. But if you are writing software today then needing absolute control over where exactly your data touches is a good reason to prefer C/C++ today. Not that that's a very common need.
I have a hunch this is why there's no Semaphore implementation in the Rust standard library, though it could be due to fundamental inconsistencies in semaphore APIs across OSs as well ¯\_(ツ)_/¯
Furthermore, they still present a fair amount of design challenges in the specific context of Rust: https://neosmart.net/blog/implementing-truly-safe-semaphores...
[0] technically they were there, added in 0.4, never stabilised, deprecated in 1.7, and removed in 1.8
Well, I do appreciate your work, and the information is certainly helpful.
But it's a bit like a urologist explaining what it will be like when you pass a kidney stone.
And then find out that the C++ standards committee is working on a new kidney-stone shape that's backwards compatible, but adds more jagged spikes.
What's your favorite "you forgot X"? You should definitely make an updated v2 because every link I've opened from the bottom half has been absolutely bonkers.
Three dimensional analog literals drawn using ASCII? What the flying hell was the standards committee thinking.
else while
And if the basis of c++ is c with classes, you’re horribly incorrect. C is a small language, easy to understand and hard to master.
Rust is fucking miserable to understand, and as such, hard to master.
Why are you being pedantic?
More seriously, the typical response to undefined behavior is for the compiler to optimize out whatever code may trigger it. It is great for performance and one of the reasons C and C++ often top the charts, but it may lead to weird bugs that depend on the optimization level.
For the union case, the compiler could completely remove the code that reads the float: because it is UB, it is something you shouldn't do, so it considers that you will never do it, so it can just remove the code. In the end your conversion function may do nothing at all (very optimized!). In practice, because it is so common, even though it is UB, it will usually do what you want it to do.
Almost every part of C++ creeps into almost every other part, and C was already complex enough... and let's just ignore that C++ is not completely compatible with C.
I think I found it via a stackexchange answer about how the "Wiring" language for Arduino sketches differs from regular c++. In Wiring, it's mostly things like no rtti, no virtual methods not resolvable at compile time, no exceptions, unsafe-math, permissive casting, pre-build system concatenates all .ino files into one .cpp file, very limited libraries, and some default includes.
From a skim: <chrono>, <filesystem>, <thread> (along with other concurrency libraries like <mutex> and <future>) are the main ones.
As far as language features that we ban, it's notable that we ban exceptions. Rvalue references are generally discouraged outside of well-trod paths (e.g. move constructors). We ban runtime type information (dynamic_cast, typeid, etc). There are some newer features in C++20 that we haven't accepted yet (modules, ranges, coroutines), due to lack of tooling, concerns about performance, and more.
Sometimes these bans get reversed (there was a long-standing ban on mutable reference parameters that was overturned after many years of discussion).
One of the key points we try to emphasize is that readability is primarily about reading other people's code, and so we provide well-trod paths that try to avoid pitfalls and generally surprising behavior (e.g. this article's focus on initialization? https://abseil.io/tips/88). There is value in moving closer to Python's "There should be one-- and preferably only one --obvious way to do it."
We assert that one of the goals of the style guide (the second section, after Background) is that rules should pull their own weight; we explicitly cite this as a reason why we don't ban goto. I imagine this is also why there isn't an explicit ban on alternative operator representations (e.g. writing `if (foo and bar) <% baz(); %>`).
I don't think I agree that every rule pulls its own weight -- I like to cite that I copy-pasted the internal version of the style guide into a Google doc earlier this year to see how long it was... and it clocked in at over 100 pages. But maybe that's an indicator of how complex C++ is, where we have to make the tradeoff between being concise, being precise, and providing sufficient context.
(I'm sure it is.)
"Is the style guide they use for internal projects the same as this style guide that they have published externally?"
"could you clarify which ones you're talking about?"
"Is the internal style guide you described the same as this one I found in google's account on github?"
"oh, I see what you mean"
will you send your second, or shall we simply pistols-at-dawn?
> Why are you being pedantic?
I'm not. I'm making a real point: Rust is much simpler and easier than C++, so the spirit of Bjarne's quote, which is that for a language to become popular it necessarily has to have as many drawbacks as C++, is wrong.
With all due respect to your expertise, the whole idea of a footgun is that it tends to go off accidentally. The more footguns a language contains, the more likely you are of accidentally firing one.
This is “I use the language I always use because I always use it”, and not actually a praise or C++ specifically.
In my 21 years coding professionally, I will settle on C++ or Ruby for most problems depending on context. Ruby for quick, dirty, and "now!", while I use C++ for Long lasting, performance, strongly typed, and compatible things. Those aren't perfect categories and there are reasons to pick other tools, but Choosing C++ after a long career does mean something more than "I am comfortable with this".
Does it?
If you are fortunate enough for your domain to have good IO libraries then there is a chance you can do everything the "Modern" way avoid a lot of the headache and avoid most of the footguns entirely. That maturity and omni-pattern availability is potent, but all that power does come with the possibility of mistakes or surprises.
Compare to newer languages where we don't know what might break or might need to do something the language omits as part of its paradigm. I have done a ton of Ruby projects and about half the time we need more performance so I need to bust out C to rewrite a hotspot in a performance sensitive way. Or sometimes you just really want a loop not a functional stream enumerator like is the default in Ruby. For a new language, Theo tried the 1 billion row challenge in Gleam and the underlying file IO was so slow the language implementers had to step in.
This is an engineering and a business choice. There are reasons to avoid C++ and footguns, like any risk, are among them. These aren't risks without mitigation, but that mitigation has a cost. Just like newer languages have reasons not to use them. A team needs to pick the tools with risks and other cons they can handle and the pros that help them solve their problem.
The problem is that your definition of risk may not be the same as others', and so there isn't always agreement on what is ok and not ok to do. And regardless, humans are notoriously bad at risk assessment.
> This is really well explored space in C++. Just using value semantics, shared_ptr, and RAII instead of naked news and reckless mallocs would improve several "old" codebase I have worked in. Maybe people shouldn't be reaching for const_cast so often, and similar.
Right, and all that is exactly the point: all of that stuff is in wide use out there, and I suspect not just in "old" code bases. So there's still not consensus on what's safe to use and what's too risky.
And regardless, I have enough to think about when I'm building something. Remembering the rules of what language features and APIs I should and shouldn't use is added noise that I don't want. Having to make quick risk assessments about particular constructs is not something I want to be doing. I'd rather just write in a safer language, and the compiler will error out if I do something that would otherwise be too risky. And as a bonus, other people are making those risk assessments up-front for me, people in a much better position than I am to do so in the first place, people who understand the consequences and trade offs better than I do.
I really like this value proposition: "if the compiler successfully compiles the code, there will be no buffer overruns or use-after-free bugs in it" (certainly there's the possibility of compiler bugs, but that's the only vector for failures here). For C++, at best, we can only say, "if you use only a particular subset of the language and standard library that the compiler will not define or enforce for you (find a third party definition or define it yourself, and then be very very careful when coding that you adhere to it, without anyone or anything checking your work), then you probably won't have any buffer overruns or use-after-free bugs." To me, that's almost worse than useless; even if I find a C++-subset definition that I think is reasonable, I'm still not really protected, because I still have to perfectly adhere to it. And even if I do, I'm still at the mercy of that subset definition being "correct".
Is a footgun a gun that only aims at feet? Because that seems like a dumb thing to invent in the first place. Or is it a gun that happens a to be aiming at feet? That seems like something that could only exist by user error.
I think “enough rope to hang yourself” is a more accurate description of almost every programming languages, since rope is at least intended to be useful (although it is a bit more morbid of an analogy).
This is a footgun - the way to avoid the holster firing is to simply not wear sunglasses, or keep something in your left pocket, and then it would never fire. But the problem is that both of those things are extremely common (for good reason). It’s a poorly thought out feature because it has severe consequences (potentially shooting your foot) for extremely common situations (wearing sunglasses and using your left pocket).
I think you're being a bit too literal. It's not an analogy at all, and this has nothing to do with firearms best practices. If we were to define a footgun as "a gun that is only capable of shooting you in the foot" (or perhaps more broadly usefully, "a gun that in theory can be useful, but it is nearly impossibly difficult to make it do anything other than shoot you in the foot"), then the entire point of using the term is to describe something that has no useful, logical purpose, and is unsafe to use, even as designed.
Being "given enough rope to hang yourself" is indeed another good idiom to use for things like this, but the implication is different, I think: when you're given enough rope to hang yourself, the outcome is still very much in your hands. You can intentionally or unintentionally use that rope to hang yourself, or you can be reasonably expected to use that rope in another way that would turn out to be safe or useful.
"Footgun", by contrast, is used to describe something that has no (or negligible) safe uses. Maybe the original intent behind the design of what's being described that way was to have safe uses, but ultimately those safe uses never really panned out, or were so dwarfed by the unsafe uses that the safe use isn't worth the thing existing in the first place. But, unfortunately, there are some people -- maybe only 0.01% of people -- who are able use it safely, and critically depend on that safe use, so we can't completely destroy all these footguns and save everyone else from the error of their ways. And unfortunately most everyone else sees these 0.01% of uses, and believes they are so useful, so efficient, so brilliant, they want to try it too... but in their hubris they end up shooting themselves in the foot, like most others before them.
> 1972 - Dennis Ritchie invents a powerful gun that shoots both forward and backward simultaneously. Not satisfied with the number of deaths and permanent maimings from that invention he invents C and Unix.
Some of us learn to lean to the side right before pulling the trigger...
http://james-iry.blogspot.com/2009/05/brief-incomplete-and-m...
It is any trap in something technical that is likely to cause problems from perceived normal use.
Compare to related terms: "Pit of failure", "Turing tarpit", and "Pit of success".
Choosing no to decide you still have made a choice.
That is why the epochs proposal was rejected.
Additionally, the main reason many of us, even C++ passionate users, reach out to C++ instead of something else, is backwards compatibility, and existing ecosystem.
When that is not required for the project at hand, we happily reach out to C#, D, Java, Go, Rust, Zig, Swift, Odin,.... instead.
None of the edition changes that Rust has made have any effect on the ABI. It also has no stable Rust ABI, so there wasn't an effort to formalize that, but 1) ABIs should always be versioned (to avoid getting stuck with a bad ABI) and 2) you can use editions for other kinds of change in the meantime.
> across compilers,
This is almost tautological. Yes, having two C++ compilers agree to their support of editions is the same as them agreeing to their support of concepts. I don't see how this is a critique of the strategy.
> template code across editions, with semantic differences.
Rust defines editions at the compilation unit level: the crate. But it has macros (which are akin to templates) and editions are handled at that boundary (you can have code in one edition invoke macros in another) because the compiler tracks editions at the token level (the information is attached to their Span). There's no reason editions in C++ can't work with templates. You would have to specify the edition of the template, and given C++ semantics you might have to have an opt-in scope to say "use this edition here, override the rest of the file", but it would be possible.
Example, there is no story for scenario, where a callback defined in one version, is used in another crate version, calling into code, using yet another version, while passing a closure with a specific type with semantic changes across all versions.
I am not yet convinced they will scale at the size of industry use cases for C and C++, with a plethora of compilers, and mixing several editions on a 30 year old codebase.
or more specifically they only work with ABI stability if they ABI doesn't change between epochs
which isn't a issue for Rust because:
- it is cleanly modularized
- it build a whole module "at once"
- it doesn't have a "stable" ABI (outside of "extern/repr C" parts which don't contain non reprC parts rust doesn't even guarantee ABI compatibility between two builds in exactly the same context*(1))
- tends to build everything from source (with caching)
- a lot of internees are intentionally kept "unstable" so that they can change at any time
on the other side due to how C/C++ build things, doesn't have clean module isolation, how it chooses build units, how all of that is combined, how it's normal to include binaries not build by your project (or even you), how such binaries contain metadata (or don't) and how too much tooling relies on this in ways which make changes hard, how it doesn't have build-in package management, how it you specify compiler options and how compiler defaults are handled etc. come together to make that impossible
in a certain way how you specify that you use C++11,17 etc. is the closest C++ can get to rust editions
like initially it might seem easy to introduce syntax braking changes (which most rust edition changes boil down to) but then you realize that build units using other editions have to be able to read the header file and the header file e.g. in context of templates can contains any kind of code and that header includes aren't that much different too copy pasting in the header and that you don't have a standard package manager which can trace which edition a header has and endless different build systems and you kinda give up
purely technically it _is fully possible to have rust like editions in C++_ but practically/organizationally in context of e.g. backward compatibility with build systems it's just way to disruptive to be practical
Which is all well and good for us, the application developers. But if C++ wants to exist in the future as a thriving language (as opposed to moving in to the nursing home with Cobol and Fortran), then it needs to come up with some solution to remove cruft.
https://en.wikipedia.org/wiki/C%2B%2B17#Removed_features
https://en.wikipedia.org/wiki/C%2B%2B20#Removed_and_deprecat...
https://en.wikipedia.org/wiki/C%2B%2B23#Removed_features_and...
Part of their ‘problem’ is that they have lots and lots of users with long-living code bases. That means that, if they move fast and break things, their users won’t move to newer versions of the language.
Another part is that they want to be able to generate the fastest code possible. That leads to such things as having all kinds of constructors (‘normal’ ones, copy constructors, move constructors), and giving developers the ability to tweak them for maximum performance.
In addition, a lot of this was invented after the language was in use for decades. I think that makes the constructor story more complicated than needed.
https://en.wikipedia.org/wiki/C--
The GNU gcc/g++ was far more important to standardization than most people like to admit.
Have a great day, =)
It is possible that the entire field of programming is full of ridiculous people. But it seems more likely that C++ is like a gun with no safety, or something along those lines.
An obvious example is omitting NULL pointer checks. Passing a NULL pointer to certain system library functions is UB even if it would ordinarily be expected to be reasonable, e.g. memset(NULL, 0, 0), so some compilers will see that you passed a pointer to memset, and passing a NULL pointer to memset is UB, therefore it can omit a subsequent NULL pointer check guarding a call to something else when the something else isn't going to behave reasonably given a NULL pointer.
This is an insane footgun, but it also allows the compiler to omit a runtime NULL pointer check, which makes the program faster, so people who care most about performance lobby to keep it.
They do! Here's a reference. https://en.wikipedia.org/wiki/Slamfire
Another common way to shoot yourself in the foot is a gun which will go off if you drop it. An example of a gun where early models were especially susceptible is the Lanchester: https://en.wikipedia.org/wiki/Lanchester_submachine_gun
It's an idiom. It's not supposed to be entirely logically consistent. It means what it means because people have decided that it means what it means. Your objections aren't really relevant.
"Footgun" is I think a fairly recent addition to the English lexicon, but it's based on the centuries-old "to shoot yourself in the foot". It seems silly to argue with centuries of English idiomatic usage; no one, by definition, is going to win an argument against that.
The fact that you put "internal" in quotes suggested to me a mild level of sarcasm or disbelief, i.e, I read your message as "You mean this style guide, published on the internet, for all to see? Clearly that's not 'internal'!"
Either way, to me, "internal style guide" (regardless of any quotes placed around any word) means "style guide that is internal" (that is, the style guide itself is private or unpublished).
But the person you were replying to called it a "style guide for internal c++ code": that word ordering makes it clear that "internal" is describing "c++ projects", and that the internal/external (or unpublished/published or private/public) status of the style guide itself is not being talked about at all.
(As an aside, if the commenter upthread had instead said "internal style guide for c++ code", that could have also meant the same thing, but would have been ambiguous, as it wouldn't have been clear if "internal" was describing "style guide" or "c++ code", or both, even. But "style guide for internal c++ code" is about as unambiguous as you can get.)
what noun cluster would you use for the style guide that google uses internally? Their internal style guide is perfectly accurate. If they publish it externally, does make it no longer their internal style guide? Nope. Would it make somebody exploring the topic wish to add the clarification "oh, their internal and external style guides are one and the same"? Yes.
None of that conflicts with what I wrote, but it does conflict with what you wrote.
Like I said, I never wrote c++ there so I’m quite likely misremembering details, but IIRC the published style guide linked to upthread is more or less a snapshot of the previously-unpublished style guide used internally for internal code. It may have some omissions and it’s quite likely out of date.
I’m just amused that this discussion about semantics reminds me so much of the gif linked from https://news.ycombinator.com/item?id=40882247
Uhm, no. I think you've got that completely backwards. Agree entirely with what they said, not at all with your interpretation. I think you're flat-out wrong here.
Whereas the respondent said this:
> Google has a “style guide” for internal c++ code
This is a style guide for definitely-internal c++ code, with the internality of the style guide itself unspecified. I'm not sure what the effect of the scare quotes around "style guide" is meant to be, just that it doesn't move the internal part next to the style guide part.
Putting the whole thing in quotes, rather than just "internal", questions whether the guide you found is the guide referred to, rather than the internal nature of the style guide itself, which the quoted sentence takes no position on.
This has been your daily dose of Hacker News style guides for discussing style guides.
Consider that the style guide that AirBnB uses for internal projects is not the same as the style guide they publish externally, and you can sympathize with why the distinction matters :P.
But if you’re going to behave like a child I’ll skip the rest. Enjoy!
and btw, the Hesperus/Phosphorus reference I made many commments ago refers to a famous example of these exact questions in the Philosophy of Language. Read up on it and you may begin to know as much about this as I did before we started.