Furthermore, Rust's type system is absolutely not only designed to prevent undefined behavior. Dereferencing a null reference in Java or C# is not undefined behavior, but it is still a (tremendously common) bug. Deferencing a null reference in Rust is not possible because null references do not exist.
Affine typing allows Rust programs to statically guarantee that only routines applicable to an object's current state can be invoked; in most other languages this can only be checked via runtime assertions.
Unique ownership not only guarantees lack of data races but allows programs to statically prevent race conditions more generally, including those that in other languages may occur even in the absence of any concurrency (see eg. Java's ConcurrentModificationException).
Rust enums and exhaustive pattern matching statically ensures that all possible values of a type are accounted for. Again, not doing so might not directly lead to undefined behavior, but would be a compliance failure nonetheless, and again, Rust has your back there.
I don't think that a program must be free of undefined behavior before significantly reducing bugs. I think most people would prefer a program with five undefined behavior bugs and no other bugs to a program with zero undefined behaviors but 50 other bugs (assuming all those bugs are of similar severity).
Also, even clearing that bar doesn't require doing so with the type system (e.g. Python doesn't have undefined behavior, either). It also doesn't mean that clearing it gets you very far with respect to functional correctness.
> Affine typing allows Rust programs to statically guarantee that only routines applicable to an object's current state can be invoked; in most other languages this can only be checked via runtime assertions.
Yes, but that doesn't mean that the end result is more correct, and, indeed, there is no evidence to suggest that it is. Software correctness is an extremely complicated topic, one that I have written extensively about [1], and just pointing out that one technique can eliminate a certain class of bugs is almost never a guarantee that that technique is superior to other techniques, even if they are partial.
If python doesn't have a type system, why does it report type errors like this?
>>> "foo" / "bar"
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
TypeError: unsupported operand type(s) for /: 'str' and 'str'Depending on the nature and surface area of the undefined behavior, you can't necessarily know that your program has "no other bugs" if you can't trust the soundness of the rules for writing your program to begin with. You'll have to test it externally with something that you can depend on the soundness of, which is just moving the same work to somewhere else.
I.e. the mere existence of undefined behavior in the formal system does not negate the soundness of the entire system. At least it doesn't in arithmetic and it doesn't in C. In fact, C is currently the language with the best formal verification tools. Of course, that is out of necessity, but it's absolutely false that the presence of undefined behavior in the language breaks the possibility of sound formal deductions.
It is true that C does make some local logical inferences more tedious, and some of that is due to undefined behavior, but that's a whole other discussion.
[1]: Deductive proofs are very rarely used in real safety critical system, and they're especially unpopular even compared to other formal methods such as sound static analysis and model checking. Those two are more commonly used in safety critical systems, but the most common assurance tool is tests.
I'm also not sure what your disagreeing with exactly. The reason all those assistive tools for C exist is because C is popular for orthogonal reasons, but also its undefined behavior is known to be a source of danger & uncertainty. That's why compilers undergo qualification for instance. The undefined behavior is characterized, catalogued, and codified to turn it into de facto defined behavior.
No, this is not true. Again, the fact that a[i] is undefined for values of i greater than or equal to the array's length does not necessarily make guarantees about the value of a[i] more difficult when it is defined, and it certainly does not impact the soundness in that case. Verification methods can -- and do -- deal with undefinedness without making it "de facto defined". Indeed, even in arithmetics 1/0 is not defined, and yet Peano arithmetics is fine. Formal logic in general does not require every expression to be defined (in the sense of having a certain specific value) in order for it to be sound.
I agree that undefined behavior is a problem, but it is not necessarily worse than other problems when correctness is concerned, and it is not true that it affects the soundness of inferences. It can certainly be tricky or confusing, though, and it is certainly a source of incompatibilities, which is why most languages try to avoid it or control it (Rust has undefined behavior inside unsafe, and Zig can dis/allow it depending on compilation level).
A whole bunch of the tests and the developer interface agreement that come with your purchase of a IEC/ISO qualified C compiler are around establishing what that instance of the compiler does with undefined behavior corners of C. The whole point is to have evidence of turning unknowns into knowns for a defined set of deployment constraints.
> Again, the fact that a[i] is undefined for values of i greater than or equal to the array's length does not necessarily make guarantees about the value of a[i] more difficult than if it were defined, and it certainly does not impact the soundness in cases where i is within bounds.
That's because you know the surface area of what's defined and what isn't already. If the behavior for a[i] were undefined for all cases of a[i], then you wouldn't be using a[i] at all because you'd have no idea what to expect about its behavior.
> I agree that undefined behavior is a problem, but it is not necessarily "worse" than other problems
You seem to be making this argument on the basis of taking for granted that you know something about what is defined already. Here's how to know that undefined behavior is worse. Say I asked you to write a program for your own pacemaker in a language where _all behavior_ is undefined. What do you do? It will be impossible to avoid killing yourself except entirely by accident, unless you resort to external validation methods.
The assurances that are possible in things you're describing are predicated on having some known a priori model for the domain of what is defined and what isn't, and the various certification/qualification methods required to use certain technologies & tools in domains where these things tend to matter focus on keeping engineers inside the bounds of what's defined and reducing the scope of what is undefined for a given configuration in-use.
Maybe, but undefined behavior still doesn't kill soundness.
> That's because you know the surface area of what's defined and what isn't already.
Sure, but in every programming language most of what you do is defined, even in C.
> Say I asked you to write a program for your own pacemaker in a language where _all behavior_ is undefined. What do you do? It will be impossible to avoid killing yourself except entirely by accident, unless you resort to external validation methods.
It's impossible to write any program in a language where all behavior is undefined. Such a language, by definition, has no semantics, and so no notion of a program. In fact, a language with no semantics of any kind is not a language at all.
> The assurances that are possible in things you're describing are predicated on having some known a priori model for the domain of what is defined and what isn't,
This is true for all programming languages.
So, it's necessary to remove a sufficient amount of undefined behavior before something is even usable in a coherent way? Which is another way of saying, you have to have defined enough behavior to be able to even get to the point of knowing whether or not the programmer is the one responsible for the "bug".
Undefined behavior seems to be necessarily worse than programmer error, since to be able to determine the latter you have to have removed some part of the former.
:shrug:, going to get back to work.
It's like asking which is "worse", saying "England is a country in Africa" (false) or "Love ate a circle" (meaningless)? Both are "bad" in their own way.
But undefined behavior can certainly be inconvenient, because it is insufficient to determine whether a sentence is "well-formed", i.e. grammatically correct, in order for it to be meaningful, as some grammatically correct sentences are nonsensical. But it's certainly not catastrophic, because mathematics has a lot of undefined sentences, and we still use it successfully.
Wrong, wrong, wrong! A race condition is an error in program logic. It can't be prevented by a type system, which is limited to local, syntax-based analysis. Race conditions can trivially occur, e.g. in channel- or message-passing-based models of concurrency, where there's nothing that could cause a data race.
def counter = 0
def increment() {
def tempcounter = null
counter.lock {
tempcounter = counter
}
tempcounter++
counter.lock {
counter = tempcounter
}
}
I think the argument is that it's inadequately typed. I.e. the program itself is correct because the semantics aren't captured properly.
If you had a type system that captured that tempcounter was derived from an earlier version of counter, then it could enforce that the assignment of counter in the second lock does not depend on a previous version of counter. Runtime enforcement would be trivial in most languages by making counter a monad that essentially carried a DAG around to detect this sort of problem. I'm pretty sure though that you could enforce this at compile time with dependent types in something like Idris.
Undefined behavior means doing something the compiler spec does not specify, therefore the compiler is free to do anything. There are plenty of programs that exploit the programmer's knowledge of a particular compiler's actual, unspecified behavior as an optimization.
One example is using an int variable after an overflow. Post-overflow, the int's value is undefined behavior, but in practice it is predictable for a given compiler and platform. It is possible to write a program that implements its own program spec correctly, while exploiting out-of-spec behavior of its compiler under the hood.
Much like how it's possible to wrap unsafe Rust code with a safe interface.
Unsafe code blocks still have defined and predictable behavior, but you have to guarantee the memory safety.
There are plenty of programs that have bugs.
Programmers who exploit undefined behavior on purpose are reckless.
And then they break when the compiler changes its behavior…
In Rust, as well as GC'd languages like TypeScript, dereferencing a null reference is possible, in the same sense, because a type cast or unsafe code may have been incorrect.
True, but type casts are unsafe code. You'll never see a null pointer so long as you stick to the safe language subset.
A program that has undefined behaviour is not correct, unless the spec has undefined behaviour (why would it?).
> On that front, there seems to be no evidence, nor even hints, that either Haskell or Rust achieve better correctness (or the same correctness more cheaply).
I don't know what you mean by evidence, but Haskell has a stronger and more expressive type system than say Java, it doesn't have null pointers, it has referential transparency, and simpler, safer concurrency and parallelism builtins. The fact that you can make more general assumptions about pieces of code which you don't necessarily fully understand makes writing safer and correct code easier.
> But Rust's type system, and certainly Haskell's -- both interesting but very expressively "weak" type systems -- have not been found to be correlated with higher correctness.
Are you saying there have been studies that show Haskell's type system has not been correlated to higher correctness than say Java or Python, or are you saying you are unaware of any such studies? Also, Haskell's type system is certainly more expressive and strong than Java's or Python's.
Yes, but a program has many ways in which it cannot be correct, and eliminating one of them does not mean that you've eliminated a significant amount of problems, and it doesn't even mean that the amount of incorrectness you've eliminated is bigger if you'd spent your time with other approaches.
> The fact that you can make more general assumptions about pieces of code which you don't necessarily fully understand makes writing safer and correct code easier.
Except this does not seem to be the case, as evidence does not suggest that Haskell programs have significantly fewer bugs or are produced faster.
> Are you saying there have been studies that show Haskell's type system has not been correlated to higher correctness than say Java or Python, or are you saying you are unaware of any such studies?
I am saying that there is no evidence to suggest Haskell programs are more correct, and that at least one study has tried to find such correlation but did not.
> Also, Haskell's type system is certainly more expressive and strong than Java's or Python's.
It is, but it is still very weak.
What evidence?
> It is, but it is still very weak.
It's strong relative to the most popular programming languages in use today.
Similar evidence to that that does not suggest eating lettuce reduces baldness.
> It's strong relative to the most popular programming languages in use today.
On an expressivity scale, as far as functional correctness is concerned, if Python is 1, Java is 2 and Agda is 100, Haskell would be at maybe 2.5. Sure, stronger, but not enough to suggest it has some real impact, and, indeed, none has so far been found.
Liquid Haskell would be a big jump (say, to 20), and anything more than that would put you close to, or at, 100.
But note that type systems are far from the only way of expressing correctness properties. Eve is untyped, but its ability to express correctness properties is much higher than Haskell's. Also, even type systems that are designed for functional correctness do not necessarily do a good job at it compared to other approaches.
Because maybe you're specifying a partial function. The code to handle the special cases of invalid inputs can sometimes have dramatic effects on code size and performance, sometimes too much to bear for code that should never run in the first place; allowing for garbage output to be produced for garbage input instead of a result that says "this is garbage input" could well be acceptable.
Furthermore, there are some places where we don't actually know how to specify the intended behavior in the first place. Data races are perhaps the best example here; we have yet to find a satisfying specification for data races that is more constraining than "anything goes," and the two most well-known attempts (Java and C++) have both been widely regarded as failures. But pointer aliasing is another area that proves surprisingly difficult: if requiring pointer provenance isn't acceptable, then trying to come up with a reasonable specification is surprisingly difficult.
That seems solvable by refactoring your types a little. Also, some languages do not allow partial functions - such as total functional programming languages.
> we have yet to find a satisfying specification for data races that is more constraining than "anything goes," and the two most well-known attempts (Java and C++) have both been widely regarded as failures.
What about nested data parallelism[1] or pure parallelism such as in Haskell?
[1] https://www.classes.cs.uchicago.edu/archive/2016/winter/3200...
I assume something analogous could happen with pointers to freed memory that are converted to integers, but that seems easier to resolve.
On the contrary! Consider for instance: "Total Haskell is Reasonable Coq"[1]
And, BTW, Coq is not a particularly effective tool for writing correct software, either. The largest software whose correctness was largely verified in Coq (CompCert) is positively tiny compared to standard software (it's less than 1/5th of jQuery), and the cost was huge. So Coq's "correctness score" of increased correctness per unit of effort is not particularly stellar compared to alternatives.
If your code relies on e.g. integer wrapping as implemented by CPU integer instructions, then your code, by design, inherently has undefined behavior for architectures that implement CPU integers using something other than two's-complement. (A lot of crypto code is invalid on these architectures!)
> GCC will optimize away the check (c < 0)
...and the suggested solution to this was to use special, third-party "safe math" libraries written in target-specific assembly.
Which meant you weren't really getting the feature we're looking for (which is portable code that will compile anywhere, using either native two's-complement wrapping arith when available, or a shimmed emulation of it when not), because you were locking yourself into a whitelist of targets that the library had been written to support.
Now, mind you, that's just C. "Using two's complement arithmetic" might be something you can get a guarantee on from, say, Ada, or Fortran, but if you can, I've never heard about it. Anyone know?
As soon as undefined behavior is guaranteed by a spec, it's no longer undefined, so your statement reduces to "you should never rely on UB." You can say that, but people will still do it because it makes their programs faster.
My first experience with running real Haskell code was installing Leksah - which is a Haskell editor written in Haskell itself.
I initially installed an older version by accident. It presented me with some window saying it will try to download something. Then it closed spewing a cryptic error message into the console. I could not restart it, because the program immediately went to the same error on startup.
I then installed the latest version. That one loaded... only to consistently crash when I tried to create a file.
This is what happens when software is written by people who don't understand the difference between theoretical "correctness" and real-life reliability.
If i had a penny for every time someone forgot to check the return code or errno in a C program i would be a millionaire by now. Such class of error is simply not possible in Rust's type system. I'd classify this error as functional correctness, it has absolutely nothing to do with undefined behavior or anything like that, hence Rust does improve functional correctness.
Other common error i see i C# for example is that because of the GC everybody just passes objects around wildly without thinking of ownership, eventually there is a reference to some object everywhere, often in multiple threads, and you have no idea who is the real owner and who is just borrowing. Every reference is mutable so it is often very difficult to understand who is mutating it. Rust's immutability by default with only one mutating reference allowed at a time, would prevent this class of error also. (This isn't necessarily an error in itself but usually this spaghetti eventually leads to functional errors)
I do see your point though and agree that it would be very interesting to see a statistical evidence of how well this translates to productivity and end user perceived functional correctness. It could for example be that all such errors are weeded out during testing anyway, which leads us down the endless rabbit hole debate of dynamic vs static programming language. Though in my experience the closer to compilation you can find an error the better. Especially in larger teams where types not only help yourself but also describe the intent of the code for others.
It isn't. Functional correctness is about ensuring that the code matches the spec. Preventing those kind of bugs could have a positive or a negative impact on functional correctness, but they are not, in themselves, functional bugs.
> Such class of error is simply not possible in Rust's type system.
This doesn't have much to do with correctness, though, and it's is a common mistake when discussing correctness (a subject I have written extensively about https://pron.github.io/). I am not saying that eliminating a certain class of bugs is bad; it might actually be good (although it might also be bad), but it's far from simple. It is absolutely not the case that if one language eliminates a certain class of bugs and another one doesn't, then the first leads to more correct programs. This implication is just wrong, both in its internal logic as well as empirically. To get some intuition for why it's wrong, consider that when naively writing a program on paper, you introduce 100 bugs. When you actually program it, you must use a variety of methods, including tests to get rid of them. Now, if you use a language that eliminates 10 of them with its type system, it does not mean that given the same amount of effort, you'll catch more bugs in that language than the other. It's possible (indeed, we know it happens when we work with more powerful type systems like dependent types) that it's harder to write the program in the first language, and in the time it took you to eliminate those 10 bugs -- i.e. getting the program to compile -- you'll have found 20 in the other language. And it's certainly possible that overall there just wouldn't be much difference.
Again, I am not saying eliminating certain classes of bugs is bad. I am saying that it is absolutely not true that doing so is a certain path to better correctness.
In any event, Rust's type system does not directly increase functional correctness. It does eliminate some bugs that could cause functional correctness issues, but whether the approach's cost/benefit is significantly better or significantly worse than, say, Java or Python is very, very far from clear.
Pointing at such papers when discussing correctness relative to other languages is like telling people they'd be much better off working at company A rather than company B, because company A gives out free lunches. Even if company B doesn't, it might offer free dinner, and even if it doesn't the salary can be higher, and even if not, it's unclear how big of a difference A's lunch is. In other words, correctness is very complicated, and there is just no way to look at one property or another and make a conclusion. Indeed, there does not seem to be a big difference in practice.
I think type systems, such as Haskell’s, strike a good balance between formal and informal specifications such as unit tests or diagrams of boxes and arrows.
Me too, which is why I say that it does not seem to make a big positive or negative difference (big differences are easy to find; there haven't been enough studies about the correlation between eating lettuce and baldness so we cannot draw any hard conclusion about there being a correlation or not, and should continue assuming that it's likely not a big cause for baldness nor a definitive cure).
> I think type systems, such as Haskell’s, strike a good balance between formal and informal specifications such as unit tests or diagrams of boxes and arrows.
I agree. Haskell's type system is OK, and I think most other languages also strike a balance that is roughly equally good, albeit not necessarily through the type system, which is why Haskell doesn't appear to be doing much worse or much better at the bottom line. So I definitely see no good reason to suggest Haskell should lead to drastically lower correctness, and if someone said they cared about correctness and want to use Haskell, I wouldn't think it's an exceptionally bad choice.
I think (simple) type systems have many good benefits (code organization, documentation, tooling, AOT-compilation performance), which is why I like them, but correctness just isn't one of them. And why should it be? Theory certainly does not suggest simple types should have a big impact on correctness. The strength of the Haskell type system is such that for most uses of it, for any well-typed Haskell program, there is a very small finite-state program that satisfies the same type (expressible as a very small program), yet clearly most of what the program does -- as well as its code -- lies well beyond its type-equivalent small FSM. So any argument that says that the type would still have a big impact of correctness despite the theory saying it shouldn't would also need to explain why other techniques, like tests without a type system, aren't at least as effective (cost/benefit-wise).
Most of those impact correctness (amongst other benefits), otherwise we wouldn't care about them. None of the studies that have been carried out tested the long-term benefits of types, which is precisely where types would shine,so I cannot agree with your conclusions in this thread.
Well, it doesn't seem like they do because, again, we can't find significant correctness differences. But I care about them because they improves my subjective experience (enjoyment). I don't know if it's universal, though. I care a great deal about the art I enjoy, but I don't make any universal claims about it improving productivity or something. And I find any such claims -- just like unsubstantiated claims about programming languages -- to be just like claims about homeopathy: I understand and sympathize with the psychological mechanism that causes people to make them, but they are unlikely to have any large real-world validity.
> None of the studies that have been carried out tested the long-term benefits of types, which is precisely where types would shine, so I cannot agree with your conclusions in this thread.
It's not a "conclusion." Until we find evidence that lettuce cures baldness -- or causes it -- we assume it doesn't.
But it's even stronger than that. We don't need difficult studies to detect large effects. Large effects are often found even without studies at all. If eating lettuce increased or decreased baldness by 50%, we would have noticed. Especially in software, which is a huge economic undertaking and therefore has powerful selective pressures, adaptive techniques have very clear adoption dynamics, of the kind we've seen with unit testing and even garbage collection. Yet we don't quite see them with Haskell. Haskell doesn't even spread rapidly inside organizations that already use it. On the other hand, it doesn't rapidly disappear, either, as we'd expect from a highly maladaptive practice. Its dynamics is similar to that of a neutral trait.
So the scientific reality is that Haskell in all likelihood does not have a large negative or positive effect on any important bottom-line metric, at least until proven otherwise. This is the current working assumption, and it's great news for Haskell fans: Haskell is probably a reasonable choice, and not significantly worse than others. Looks like the FDA would call consider it a safe nutritional supplement.
Like I said, none of the studies conducted so far actually measure this in any realistic fashion. If type consistency, documentation, code organization, and tooling had no impact on the functional correctness of the resulting programs, we wouldn't bother with any of it.
> It's not a "conclusion." Until we find evidence that lettuce cures baldness -- or causes it -- we assume it doesn't.
No, we assume nothing. Why would you even bother making assumptions? Certainly you can advance a hypothesis, but that's not an assumption.
> Especially in software, which is a huge economic undertaking and therefore has powerful selective pressures
I disagree. Software development is a comparatively meager economic undertaking compared to most other human endeavours, as you need only a computer and a human brain.
Furthermore, the selective pressures you speak of aren't really driven by functional correctness ("worse is better", first mover advantage), so I frankly expect to see little correlation between economic success of a software company and the functional correctness of its software products, which in fact seems to be the case. There are plenty of reasons why the effects of types may not be visible in the current data.
> So the scientific reality is that Haskell in all likelihood does not have a large negative or positive effect on any important bottom-line metric, at least until proven otherwise.
Haskell has plenty of other baroque features that might mask the benefits of type checking. We're talking about types in general, not Haskell specifically.
That doesn't change the fact that our current working hypothesis is that there is no large positive or negative impact. And, again, studies are not required to find large impact.
> If type consistency, documentation, code organization, and tooling had no impact on the functional correctness of the resulting programs, we wouldn't bother with any of it.
I don't understand the logical deduction here. Is functional correctness the only metric we care about?
> No, we assume nothing.
We certainly don't assume that Haskell has any correlation with increased correctness, something the article does.
Also, assuming the null effect as a working hypothesis is practically the same as assuming nothing. In either case, if we want to make a choice (e.g. on correctness) we cannot take the variable (e.g. language) into account. Similar to how you don't consider the effect of lettuce on baldness, even though nothing has conclusively determined either a positive or negative effect.
We never treat the existence and nonexistence of effects as symmetrical in practice, or we wouldn't be able to make any decision. That's what we mean when we say that if there's no evidence for an effect the working hypothesis, i.e. what we use to make decisions, is that there isn't one. Our working hypothesis is, then, that Haskell does not have a significant impact on correctness, and so the article is wrong.
> Furthermore, the selective pressures you speak of aren't really driven by functional correctness ("worse is better", first mover advantage), so I frankly expect to see little correlation between economic success of a software company and the functional correctness of its software products, which in fact seems to be the case.
Wait, but before that you said that if there was no impact on functional correctness then we wouldn't do what we do. You can't have it both ways. I say something different: Haskell's adoption dynamics show that it does not have a large impact overall on the bottom line that the industry cares about. Now, you can take back what you said before functional correctness being why we do things, but even then some measure of functional correctness is clearly important to the software industry, so we can say that Haskell does not seem to make achieving the level of correctness the industry cares about significantly harder or easier than other languages in common use, or that if it makes this harder then it makes other things easier, and if it makes it easier then it makes other things harder.
> We're talking about types in general, not Haskell specifically.
Types have a very wide range, as do non-type-based software assurance methods. I think we need to be more specific. But also remember that we're talking about comparisons, so it's not enough to say that types achieve X, but that types achieve X more effectively than other ways of achieving X. X > 3 is no evidence whatsoever in favor of X > Y without additional information.
They are if the impact only shows up over time, for instance, because it's cumulative. Smoking unquestionably has the largest impact on your chance of developing lung cancer, but it still took decades of careful study to demonstrate that causation beyond a reasonable doubt. Nowhere near this level of study has been applied to the benefits of static typing (really, no question in CS has enjoyed a comparable level of empirical study, which is a shame).
>> Furthermore, the selective pressures you speak of aren't really driven by functional correctness ("worse is better", first mover advantage), so I frankly expect to see little correlation between economic success of a software company and the functional correctness of its software products, which in fact seems to be the case.
> Wait, but before that you said that if there was no impact on functional correctness then we wouldn't do what we do.
There's no inconsistency between those two assertions. As an overly-reductive example, if my company's revenue consists of 70% inertia from first-mover advantage and 30% from having a product with fewer bugs than my competitors, then functional correctness is not the primary source of revenue. As long as I satisfy the minimal viable product threshold, inertia will keep me alive.
Even so, as professionals we presumably ought to be motivated to minimize our contribution to operational costs and maximize our contribution to revenue. If improved code organization, improved documentation, and improved tooling had literally no impact on functional correctness, then we'd be wasting considerable resources for no observable benefit. There's some slack in there to make your employees's live more enjoyable, but not as much as we focus on these other qualities.
Not really. The correlation was immediately seen once they looked because the effect was so pronounced [1]. Then it took about seven years when virtually every study showed conclusive results for it to become consensus. This is not the case here.
> Nowhere near this level of study has been applied to the benefits of static typing
First, typing isn't one thing, which does complicate matters a bit. Second, various studies came up with either null result, inconclusive results, or small-effect results. This is not a smoking situation.
> If improved code organization, improved documentation, and improved tooling had literally no impact on functional correctness, then we'd be wasting considerable resources for no observable benefit.
Why? Maybe it benefits development speed? Why did you decide that developers do things for many reasons, but these things just for correctness? And anyway, it's not like types make documentation and organization redundant, they just make them somewhat easier for me, probably because I'm not used to working with untyped languages.
In any event, any claims of correlation between programming language choice and correctness are certainly unsubstantiated, and people should stop repeating what is, at least currently, a myth.
[1]: https://www.theguardian.com/news/2005/jun/02/thisweeksscienc...
You keep saying that but everyone knows the foundational work here is in pickles, not lettuce:
https://web.archive.org/web/20190324004223/http://jir.com/pi...
Java does not enforce thread safety at compile time. I would take Rust over java just for that.
Disclaimer - work heavily in Java by day, and have written some cli tools in Rust.
For example: http://hackage.haskell.org/package/liboleg-2010.1.10.0/docs/...
One can implement linear types and Rust's borrow checker in the Haskell as a DSL. That's main difference - in Haskell, you describe a system where different parts have different properties and use type checked DSLs to express them and type checker to ensure that everything goes along well.
Safety is inherently a part of accurately implementing a program's specification, even if it's not always obvious. For example, a 1-in-a-billion race condition can invalidate the functional correctness of a program, and (Type|Memory) safety is one piece of preventing these types of errors.
Moreover, such safety is not at all what is normally considered "correctness" in the field of software correctness; the term normally refers to functional correctness, i.e. that the program satisfies the specification, not that it doesn't have undefined behavior. On that front, there seems to be no evidence, nor even hints, that either Haskell or Rust do better, or worse, than other languages developed in recent decades. It's a myth, albeit one that's often repeated in some circles.
That's not to say that type systems cannot be hugely valuable, and sometimes even necessary. For Rust, the type system is what ensures undefined-behavior-safety. In other languages it has many other advantages as well. Some type systems are even intended for correctness. But Rust's type system, and certainly Haskell's are not. Both are interesting though expressively "weak" type systems, of the kind that has not been found to be correlated with higher correctness. So focusing on that as the point of comparison completely misses the mark.