My best litmus test these days is support for multidimensional arrays because it's always needed at some point in general purpose languages. CL and Ada had it right from the start while C++ needed C++23/26 to get std::mdspan and we still need to wrap it to pass the underlying/owned memory pool around (https://rosettacode.org/wiki/Multi-dimensional_array for more).
If I want a 1000x1000 array, representing it physically as a single 1000000-element array requires one allocation, and processing it element-by-element (assuming it's stored in the same order we're iterating over it) is sequential in memory and therefore very efficient.
Representing it as 1000 separate 1000-element arrays requires 1000 allocations, and pointer-chasing every time we move from one row to the next.
Otherwise you would have an array of pointers to arrays. The usage (syntax) for them would be the same but the performance would not be.
They also have different uses. You would expect an array of arrays to be an array of arrays which share the same length. For an array of pointers to an array you would expect dynamic length arrays contained within the original array.
Even in c++ could you not just define some int [1000][1000]foo? I've never really used C++ but my C knowledge assumption is that is 1000000 continuous elements.
std::array<std::array<T, N>, M> data;
int data[M][N];
Edit:
For the stack at least. On the heap, you'd need to use a single std::vector<int> and do the indices manually, or use mdspan
It works fine in C though, or FORTRAN, or Ada, or ALGOL 60, ...
NVidia has pivoted to design CUDA hardware with focus on C++ back in , and seems to be doing quite well for them.
CppCon 2017: "Designing (New) C++ Hardware”
https://www.youtube.com/watch?v=86seb-iZCnI
They were also the ones sponsoring the ISO work on mdspan, while HPC research labs are pushing for linalg on top.
I would rather be using Ada today, but that isn't how the world moves.
I respect Ada but I would not want to use it. But I have a choice between C++ with hmdspan and C99's arrays, I choose C99 any time.
The std:mdspan proposal came from NVidia employees, alongside AMD and HPC research labs.
https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2022/p00...
Yeah, I remember discussions on comp.lang.c calling programming with Ada, or even Modula-2, programming with straightjacket.
Meanwhile governments and national security bodies got another point of view.
You mean the C99 arrays Google paid the work to clean from the Linux kernel?
The kernel removed VLAs, I am more talking about vm types. But even for VLA - while I had a small role in that undertaking myself - I think it was a stupid mistake from a security point of view to remove VLAs from the kernel. Google pays for a lot of nonsense...
You can even use unconstrained arrays to provide the same functionality that Optional does in functional-programming, provided the element-type can be an element of an array:
-- Here we define an index-type with one value.
Subtype Boolean_Index is Boolean range True..True;
-- And here we define an array indexed by it, but can also have length 0.
Type Optional(Boolean_Index range <>) of Element;
If it fits on the stack, yes.
Typical code using MD-arrays is scientific code, and the data they manipulate generally do not fit there.
If you try to allocate 10MB on the stack, that's the dev problem if the program fails, it's not the compiler job to guesstimate whether something will fit there or not (and it's impossible anyway, the compiler can't know all the stack sizes a program will ever run on).
I would very much prefer scheme if the different implementations had a working standard. But I can't take my Chez-scheme code and throw it into Guile-scheme.
But pretty good chance I can take my ECL code and throw it into SBCL or LispWorks.
Bah, I think this debate was already old when I first saw people arguing it on comp.lang.lisp in the 90s. I don't have a dog in this fight other than to reject the notion that Common Lisp is "coherent" and not "organically grown".
The original Scheme belongs in the category of languages like Standard ML and SmallTalk, where a small, careful, and talented group designed them with focus. Common Lisp seems like a bunch of smart people with competing interest and legacy baselines tried to meet in the middle. To the extent CL is more pragmatic, it's another example of "Worse is Better".
When I started building a Lisp-based machine learning framework, Guile seemed like the right choice because it provides GOOPS and generic functions, yet I still ended up with a lot of boilerplate to compensate for the lack of a strong type system.
Scheme feels to me like C is to C++: not ergonomic for large-scale application development. Go is one of those languages that has both minimalism and productivity.
-- Greenspun's tenth rule
He had some lack of conviction to scope it so narrowly.
Hopefully next they can add some error handling syntax and controls.
https://elixir-lang.org/blog/2023/06/22/type-system-updates-...
As an aside - D, Zig, Rust, even typescript got most of the lessons learned from C right
Zig has the (in)famous "Writergate": https://github.com/ziglang/zig/pull/24329
And besides Rust's high count of RFCs, there are things like async (I'm not complaining about it, but its an obvious large-scale "change"), module system changes, etc.
(To be clear, I like both languages a lot. But I wouldn't call them slow moving or right from the start.)
I like CL, but I can't agree that a stdlib that doesn't even have a string split function is batteries-included.
Objective-C in contrast was a very few additions thoughtfully added that composed cleanly and freed the programmer to actually get things done.
The hard part about making a language is creating the stdlib and tooling and support for the language, but actually creating a language itself that has more features and better features than go can be done by a single person in a few months or a year probably, depending on how much experience they have.
Generics specifically are a great example here. A single person can implement a language with go-level generics fairly easily.
A good example of where they're kind of stuck is date formatting - it's stupid, unclear, and likely a mistake, but it's not a fundamental flaw; it's just a quirk.
The trouble is that Rust is older than Go and it was already confusing people into thinking enums and sum types are the same thing, so by using slightly different syntax, iota, Go avoided the whole confusion of users thinking that enums would behave like sum types instead of actual enums.
Is your attempt at making a point that not having sum types is the massive flaw? Sum types are a useful construct, to be sure, but there are plenty of good languages without them. That's more on the design quirk end, realistically.
iota is a massive kneecap _because_ it's semantically identical to enum in C and Typescript.
> Is your argument actually that not having sum types is the massive flaw? Sum types are a useful construct, to be sure, but there are plenty of good languages without them. That's more on the design quirk end, realistically.
In a dream world sure we'd have full blown sum types (and that would give a result type which would also solve a lot of the nil-interface-combined-with-error-handling issues that I've ran into when working with go), but I can forgive that. The problem is this - https://www.zarl.dev/posts/enums
The only case I see made in there is that it doesn't like how Go implicitly converts consts. While that may be a reasonable criticism, it doesn't have anything to do with iota. It is related to the type system and applies in general. Consider the same problem exhibited here:
type Email string
func Send(email Email)
func() { Send("invalid") } // Converted string const does not satisfy Email type expectations
It was made abundantly clear when Go was released that it was intended to "feel like a dynamically-typed language". Being able to pass arbitrary values is perfectly in line with a dynamically-typed language. Realistically, the type system in Go is there to give the compiler optimization hints, not to offer type safety. Go was targeted at those wanting to use Python, without the programs being painfully slow to run. How much of a kneecap is implicit type conversion, really, when it is already in line with what the target audience is accustomed to? It is a quirk at best.
If I google this quote a comment from you comes up here on this exact topic, where you seem to have completely missed the point there too. If I link to the docs [0], the full quote is "It's a fast, statically typed, compiled language that feels like a dynamically typed, interpreted language. " So it is a statically typed language first and foremost. If you want to rehash the discussion and tell people that a flawed type system that people have been asking for a solution to for close to a decade [1] you can just re-read the last time the arugments were made as I don't think I'm going to make any headway there.
[0] https://go.dev/doc/ [1] https://github.com/golang/go/issues/19814
Right, because that is primarily how it makes things fast. Python is slow largely because it spends an inordinate amount of time trying to figure out what things are. Go knows what things are at compile time because the static type system tells it what things are and thus doesn't have to waste runtime compute on figuring out what things are, aside from when you use the reflect package, like Python does. That was its value-add — that it is kind of like Python, but faster. We already went over this...
> If I link to the docs [0]
I said original announcement, so I'm not sure why you wouldn't look there? Trying to be obtuse on purpose? Regardless, performance was indeed considered more important than being dynamically-typed. After all, if performance wasn't a concern then you'd just use Python. Go exists only because it was solving a problem that wasn't already solved. Slow Python was already solved. Type safety was already solved. It didn't need to go into those territories.
> Rust is older than Go and it was already confusing people into thinking enums and sum types are the same thing
Of course the social landscape depends on people actually using it. None of the people who weren’t using Rust at the time were magically confused about enums and sum types by the mere existence of some new and experimental language.
Rust barely existed at the time Go was first being developed. And given the history of Go and the notoriety of its core team for flatly ignoring prior work in programming languages, it’s extremely unlikely that Pike et al gave more than a cursory glance to what nascent Rust was doing at the time.
But even if they had, to suggest that they intentionally replicated a dumb thing from C but gave it a different name to avoid users being confused by a different thing from a language that roughly nobody knew about at the time is bananas.
That's nonsense. Brainfuck has shaped the social landscape despite effectively nobody using it, and absolutely nobody using it for any real work. The social landscape is not at all dependent on use.
> And given the history of Go and the notoriety of its core team for flatly ignoring prior work in programming languages
Huh? Go comes straight out of prior work. It is nearly indistinguishable from Alef. What the Go language flatly ignored was being innovative. Reasonably so, of course. It wasn't trying to innovate in programming languages so that we'd have another to throw on the heap of languages nobody uses. It was trying to solve a specific business problem using well-established methods.
So is nil. Care to make the same argument?
There is a stronger case to be made for the other two. Calling GOPATH a design mistake is a stretch as it was perfectly suited to use within Google, but it didn't fit the typical solo developer's environment. Lack of generics made writing certain types of code difficult. You could be convincing in suggesting that Go did end up being used less than it otherwise would have because of those choices.
iota? It's just a construct that generates numbers (an enum). How does that kneecap anything? If it really bothers you, you can manually number the values by hand instead. Why would anyone reject a language because it allows you to optionally choose to have the compiler assign numbers automatically instead of forcing you to do it manually? The answer is nobody. In fact, most popular languages have something equivalent to iota.
Thus “go borrowed it from C, therefore it can’t have been a mistake” is a pretty lame take. The whole point of a new language is to make improvements on what’s out there already. Go missed an opportunity to fix one of C’s most notorious mistakes. So yes, they kneecapped themselves by forcing all of the users of Go to continue dealing with this well-known footgun.
Does it mean Go isn’t popular? Of course not. C was popular. PHP was popular. JavaScript is popular. Go is popular. This is always in spite of their faults. But Go could have been better.
You're almost there, but it is wildly considered to be massive mistake in context of arrays. C has weird array semantics that are confusing and hard to get right, even for seasoned developers. That is where NULL comes to bite people time and time again. Go did not inherit C's arrays. Neither did Javascript. They go out of their way to avoid what C did. In Go, you can come close to the same semantics if you use the unsafe package, but take note the name.
Yes, they still have nil, but the scope is tightly constrained and while it is technically possible to misuse, you have to try pretty hard to do so. There are many other things that are more likely to end up being misused. Those would be the more massive mistakes.
> But Go could have been better.
Obviously. Every language ever created can be better.
From Tony Hoare: "I call it my billion-dollar mistake. It was the invention of the null reference in 1965. At that time, I was designing the first comprehensive type system for references in an object-oriented language (ALGOL W). My goal was to ensure that all use of references should be absolutely safe, with checking performed automatically by the compiler. But I couldn't resist the temptation to put in a null reference, simply because it was so easy to implement. This has led to innumerable errors, vulnerabilities, and system crashes, which have probably caused a billion dollars of pain and damage in the last forty years."
Odd that he wouldn't mention the word "array" anywhere in that quote.
> Yes, they still have nil, but the scope is tightly constrained and while it is technically possible to misuse, you have to try pretty hard to do so.
Like by not remembering to check if a pointer is nil? Or actually comparing one typed nil to a different typed nil?
> Obviously. Every language ever created can be better.
Only one of us is arguing that reimplementing C’s mistakes is actually a good thing.
Rule of thumb is to not introduce values that aren't valid. Equally, don't put in -1 for an age value, even if the language allows you to. You might later forget to validate that the age value is valid.
Yes, mistakes are possible, but these types of mistakes are pretty rare. There is some value in eliminating the possibility of those mistakes, sure, but we cannot pretend that it comes cost-free. There is good reason why almost nobody uses Rocq.
> Or actually comparing one typed nil to a different typed nil?
This is something that is likely to confuse, but not a facet of nil. It is related to interfaces. Let's not flail around like one of those wacky blow up things at the used auto lot.
Hey, at least we can now understand why you have such a hard time with nil, so that's something.
It's also worth noting that the vulnerability you link to is a denial of service vulnerability. While yes, this is technically a vulnerability, it's not the sort of thing that people are usually worrying about in the context of null pointers. If features that give rise to DoS vulnerabilities are the ones to worry about, top of the list would have to be backtracking regex engines: https://owasp.org/www-community/attacks/Regular_expression_D...
The problem is the implicitness of it and how easy it is to produce them and forget to check for them.
If it came out of anywhere else, it might have struggled even to hit the homepage here.
0: https://github.com/carbon-language/carbon-lang#project-statu...
Carbon is by its own admittance not ready to use and I think mostly relegated to solving Google’s problems with C++ right now.
Both of them didn’t ship with a standard library as robust as Go’s.
One thing that made Go popular out of the gate is it is extremely good fast to build out robust HTTP services and infrastructure.
This is a very common use case and they tailored Go to be a great fit for it. You can build your entire backend without a single third party module if desired using Go’s standard library and it isn’t terribly complicated to do so.
Of course adding generics is not something that every language needs to do. Scripting languages like Ruby don't really need this style of generics. It doesn't fit the design of the language, and it's not even clear what that would look like in Ruby.
But static typing with generics does solve a recurring problem, and we've seen some real convergence towards type hints and type systems even in staunchly dynamic scripting languages. Modern Javascript is now mostly Typescript, and they've successfully retrofitted a very advanced type system in the last place I would have expected 20 years ago.
They added enums, they added sealed classes. They're trying to get rid of null (apparently it's really hard). The problem is that in 2012, when go 1.0 was released, this should have been obvious to everyone.
Here's a famous discussion from 2009, three years before the 1.0 release (tldr: facepalm)
Sum types I didn’t really miss, because you can implement a type-safe equivalent using the Visitor pattern, and retain an interface-implementation separation that native sum types typically don’t provide.
Every compiled JAR out there has to keep working as always on a JVM with updated semantics, and worse code has to be compatible, when passing class instances around between old and new code.
Then there are the guest languages on the JVM as well.