Friends of mine that are contract programmers have been ramping on Rust for the past 2-3 years claiming the number of requests for it has been increasing exponentially. I installed it two weeks ago and have been getting a feel for it. It's like if C++, C#, and npm had an offspring. I always hoped Java would make a comeback, but Rust is looking like the real deal.
These insights are a snapshot, not a trend line. Java can be in rapid decline and still be at the top of the list. Changing acceleration, not position, is how you make a comeback.
Java is currently losing Android devs to Kotlin. That could cause a precipitous drop in direct usage in the coming years.
Despite some surface differences, Kotlin and Java are so similar that "losing" one to the other isn't really that big a change. They run on the same platforms and they use the same frameworks, dependencies, debuggers, and other tooling.
A Java developer will pick up Kotlin much faster than, say, a Postgres developer will pick up Mongo. It takes a lot longer for a Java programmer to learn the Android environment than the Kotlin language.
If there were a trend from Java towards Rust or Go or Python, that would be a significant change in the world. If Java were to vanish tomorrow and everybody had to pick up Kotlin, the world would be pretty much the same in a week. Kotlin is nice, benefitting from a clean-sheet implementation of lessons that Java takes on only with difficulty, but it's not a wholesale break.
They are tools used by programmers. They do, in fact, have languages of their own, but more to the point they have whole ecosystems of expected behaviors, idioms, best practices, etc etc etc. Those take longer to learn than a programming language.
Similarly, "web app" isn't a programming language, and neither is "server side". But they are toolkits that a developer must learn on top of their programming language, and it takes longer than the actual language itself.
True, but I don't think Java is declining, I just think Python and JavaScript grew beyond Java. But I really doubt that the use of JavaScript and Python that took on replaced the use of Java, I think it's just new use cases. I haven't looked at the data, but ya, I'd doubt Java is declining.
P.S.: Also I say Java, but I mean JVM, I club all JVM languages together, since that's more what I care about, is the strength of the Java ecosystem.
And in the cloud native world everything is on the Go track. Java were huge for big enterprise monoliths, but it seems like this is not what we are heading for in the future.
I certainly agree that language rankings don’t really matter. In very broad strokes, “used everywhere,” “used many places,” “used some places,” “pretty niche,” “academic only,” these kinds of things are more useful than arguing which is 5 vs 6.
Exactly. Sometimes it takes more developers to maintain code than to initially implement it, which means technology choices today may lead to bigger popularity trends down the road.
In the same way Rust appears in ... (5,1% - 4.8% Professional Developers). The big players are looking for commercial long term safety about the used tools, platforms, programming languages and their ecosystems. After that, the technical discussion will take place.
I'm not sure who "the big players" are in your book but the largest software engineering organizations in the world are already using Rust. The post is about one of them committing their resources to the project itself.
Caution doesn't make money, it just prevents losing as much money. When it comes to technology if you're waiting for everyone else to validate an idea, you're too late.
"...the largest software engineering organizations in the world are already using Rust ...", and also for a long long time have been using Java, JavaScript, SQL, and much more programming languages. ¿It's only about new (and better) ways to develop software or can also be 'cool' to interfere in the evolution of the programming language that you plan to use for the next years ?
I 'm not against to use Rust or any other programming language, open source project or technology. I mean, the companies love open source projects because they can produce their self flavor as a product (search engine, linux distro, containers and orchestration technologies, etc.) when they wish that, without worries about uncomfortable dependencies.
AWS loves Rust in the same way than Facebook loved PHP.
No not at all but even if that we’re the case fb taking an interest spurred PHP from a legacy language back into a living language that’s growing in features.
> the largest software engineering organizations in the world are already using Rust.
That doesn't mean much because it is true for probably most serious technologies out there. Most of those organizations also have a perl script somewhere. Does that mean their core products are written in perl? Nope.
That being said, Rust is certainly successful and seeing growing adoption.
We use rust for everything at MobileCoin. I like to think of the compiler as fussy, but when you make the compiler happy your code just works. No buffer overflows as long as you don’t use unsafe code, multi-threading by default. The audits on our code have basically only turned up logic errors and out of date crates (yes it’s always possible stuff was missed, but it feels a lot different than the C/++ and functional audits I’ve been involved with in the past).
We love it and wouldn’t trade it for a different language at this point.
P.S. we’re hiring remote and sf Bay Area. If you are interested in working on digital cash for use in messaging applications, drop me at line!
I don't love Java like I love Clojure but I think Java may stage a comeback with GraalVM: it has cut down startup times to almost match with native C/C++ . I am taking all my java projects meant to run on AWS Lambda and am running it through GraalVM to cut startups time and also memory size.
I don't think Java ever left. It's still used everywhere I've worked in the past 15 years, usually quite extensively. Just because it's not the hot language to talk about on HN it's still out there and quite healthy. Personally I'd been hoping that c# and it's kin would take all of Java's core business, but it really hasn't
I don't think Java is going anywhere. It's more performant than JS and especially Python, the other 2 top 3 languages. And unlike Cxx/Rust, you have the convenience of GC. And the new GC's are VERY good.
It has a nice niche where it's less chaotic than JS and has better threading support and uses less cpu/ram than JS and python. But it's still high level enough to not be inconvenient to work with. These traits make it a great fit for beefy back end machines at big companies.
Rust will always be targeted at low level work unless they add a GC. Even with lifetimes etc a GC is just way more convenient
Applications of Java and Rust do not overlap. As a big data engineer, Rust does not provide any advantage to me. Good luck rewritting all the big data tools from Java (Elasticsearch, Spark, Kafka, Hadoop, Neo4j, Deeplearning4j, Cassandra, Solr, Arrow, OrientDB) and all the scientific stuff from C++ (Eigen, Tensorflow, PyTorch, Jax, Halide, OpenCV and the multitude of resources for CUDA and OpenCL) and databases (MySQL, MongoDB). If I want to contribute to LLVM or JVM, they are C++, the browsers are all C++. Unreal Engine and Unity (IL2CPP and engine) are all C++. Are you doing microVMs or drivers? Then probably Rust is a good investment (discutable), otherwise, I would wait for better alternatives. I will stick to Java, Scala, Python and C++ for the foreseeable future.
EDIT: may the downvotes come from the idealists that hammer down pragmatists, like our opinion doesn't even count
If by "hijack" you mean offering a contrarian opinion then this site would be rendered useless. This is a puff marketing piece for AWS first and Rust second, you cannot expect 100% fawning adoration (dont worry, there will be plenty still).
If by "you appear" you wanted to say "it appears to me" OK. It didnt appear to me.
OP is a little big negative, so that's great to temper the "a little too positive" comments.So it's a win in my book.
Yeah I agree with your last point, but I also consider the ones who downvote because that equally as petty, adding more noise to the site.
My comment in on point, not off-topic. I had no intention to "hijack" the top comment, didn't even think about that until I've seen your comment. I don't see Rust as a replacement for Java and I also pointed out that Rust has to win the mindshare of C++ programmers first, before even reaching other ecosystems. Rust is mainly a replacement for C++.
Your comment that Java and Rust don't overlap is valid, but going on to discuss how you're going to keep using C++ and Python, and how Unreal Engine uses C++, is wildly irrelevant.
Rust isn't going to replace Java, but it may supplant Java in a lot of fields where one wants both memory safety and an OO language with high performance. No one is (rationally) suggesting that Rust will replace every language.
The tone of your post, however, makes it sound as though you have something against Rust, and that you're more interested in pointing out all the things it's not good at, or shouldn't be used for, rather than contributing to a constructive discussion.
> I also pointed out that Rust has to win the mindshare of C++ programmers first, before even reaching other ecosystems
Not entirely true; Rust is an alternative to C++, but it's vastly more accessible to people who've used other languages like Java, Python, Ruby, and so on, where you get types, OO, and memory safety. Instead of simply resulting in a 1:1 replacement of C++ with Rust, it may result in a minimal reduction in C++ programming and a notable increase in Rust programming which would have been done in another safe OO language instead, but at dramatically worse performance.
Small clarification — Rust very deliberately does not do OO. It has features that look sort of like OO from a syntactic standpoint, but being object oriented is not a design goal for Rust.
> I also pointed out that Rust has to win the mindshare of C++ programmers first, before even reaching other ecosystems
C++ is just a tool, as is Rust. Rust does not need to win the mindshare of C++ programmers first to become successful. It only needs to itself as a viable alternative in whatever fields people find it useful. I believe it has.
Between WASM, embedded, and even Web development, I see a lot of "mindshare" being attracted to Rust.
Coffee Script is a transpiled language and it has become irrelevant when lot of the language features it offered is now natively available in ES6 and up (it has definitely inspired lot of ES6 features like fat arrow functions).
Rust on the other hand is not a thin layer on top of something like Coffee script. So it does not need to keep up with 'something' for decades. As long as it continues to do its current job of making systems programming safer and fun, it is going to be here for a long long time.
I don't think the transpilation was the issue for Coffee Script. After all, most usage of ES6 is through a transpiler anyways. Rather, I think ES6 is simply a superior language to Coffee Script. That's why it has lost marketshare. Likewise, something superior to Rust could arise, and then Rust's marketshare will be in trouble.
"Good luck with" is slightly negative, but overall I don't see the post as having the traits you describe.
They're replying to a line about a Java comeback by saying there's a lot of big data libraries/systems used by Java developers that it'll be difficult to compete with.
I don't know the big data space well enough to judge.
Both are general purpose languages, so of course they overlap.
> Good luck rewritting all the big data tools from Java (Elasticsearch, Spark, Kafka, Hadoop, Neo4j, Deeplearning4j, Cassandra, Solr, Arrow, OrientDB)
Cassandra (https://scylladb.com) and Kafka (https://vectorized.io) have already been rewritten once in C++, with massive latency and throughput improvements. No reason why they couldn't get their superior Rust clones in the observable future.
It is not a single dimension problem as they try to frame it. Performance is a single dimension of systems. Java with GC is a memory safe solution compare to C++ (hello segmentation faults). The reason why the C++ implementation is faster is not that it is written in C++ but it is written using a different approach. There are plenty of Java HFT systems that run circles around C++ systems yet we cannot draw a conclusion that Java is better suited than C++ for HFT. Implementation details matter, code quality matters and performance does not come purely because you pick a particular programming language. Btw. most of the systems which I was working on using Cassandra instead of ScyllaDB because the companies were not ok switch up for some performance gains (because Cassandra was performant enough for their business use case). I think what you receive as superiority is not as clear cut as you try to make it.
Debatable. Java's dominance in the data engineering space rests a bit on clay feet: it's unsuitable for scientific computing in a number of ways: for example you can neither write fast nor clean numerical code in java and it does not interface naturally with the languages that people do write numerical code in. This is not an absolute deal breaker, but it adds friction. Also, serveral big data tools like cassandra (scylladb), kafka (redpanda) etc. have basically already been rewritten in C++ not least because they suffer from java-typical performance problems (even if those problems are not necessarily inherent in the choice of java). But C++ is not a very attractive language to write code in if there are viable alternatives, because it is objectively unpleasant and expensive to use (long compile times, terrible build and packaging eco-system, high cognitive load, difficulty of avoiding undefined behavior, etc.). Rust shares some of these defects and has some of its own, but it also offers substantial benefits for some types of problems that seem relevant to data engineering and crucially it interfaces more naturally with Fortran, C++ and Python than Java does. So I think it has a chance to gradually make inroads.
> C++ is not a very attractive language to write code in if there are viable alternatives, because it is objectively unpleasant and expensive to use
From a current perspective I may agree, but in the ‘90 a number of programmers like me was really excited to be able to develop at a much higher abstraction level than C using classes and, later, templates, preserving efficiency and performances.
Yes, Java applications and Rust mostly don't overlap, but there are exceptions in both ecosystems.
Rust might offer some advantages for Data Engineers (the same way as CUDA applications in C++), not all of them as with every language else, and yes you can write Data centric libraries in Rust for Python (easier than C/C++ IIRC).
And no, there's no need to rewrite existing services to Rust, it can interface with all the examples you cited.
To me Rust is just a simpler and safer C++, if you said you would program C++, you probably will want to program Rust, writing C++ bindings for Rust is basic stuff nowadays
There are a lot of database-like services written Java (Kafka, Cassandra, ElasticSearch, Neo4J, HBase) that would likely be better in Rust. Yes the common denominator is Apache but they seem to have bought into Rust as well https://arrow.apache.org/blog/2019/02/04/datafusion-donation...
Your comment seems to assume that your use case “big data engineer” is the only use case or the only Java use case and that the only context in which Rust can grow is to rewrite the existing tools that you use today (as opposed to greenfield projects like firecracker or bottle rocket). Your edit also writes Rust off as “idealistic” and not “pragmatic” without support, which is very far from the impression I have (based on my usage and seeing it used in real applications). In general, your comment seems like generic programming language flamebait.
Firecracker and Bottlerocket are fine pieces of engineering. If Rust wants to get a share of the pie when it comes to data engineering, then it needs to have at least a killer app with an ecosystem around it. We can talk all day and all night for why some languages are more successful than others but it seems like ecosystems are giving them a boost. Hopefully, Rust will be the language for microVMs and kernel-level modules. I was strictly referring to people that think Rust would displace Java as idealists.
I generally agree--languages do better when they become the language of a particular ecosystem and that ecosystem becomes relevant. Go had this with containers and network software a bit more broadly.
That said, I don't think anyone was suggesting Rust would displace Java. The OP was pretty clear IMO that his firm had experienced exponential growth in interest in Rust skills while he had been hoping that interest in Java would grow instead. This doesn't mean that Rust will replace Java for data engineering or any other domain in particular.
I can't absolutely say that this won't happen, but given Rust's focus on writing "correct" code, e.g. the borrow checker, etc, it's really unlikely given that one of the largest areas where Python and R shine is in exploratory data analysis. I can't imagine trying to play around with unstructured exploration of a data set in Rust's (totally reasonable) syntactic verbosity, correctness straitjackets, and compilation time.
Some of that complexity can be abstracted away with a good framework. I think it’s almost inevitable that a Pandas-like framework will eventually emerge for Rust (probably based on Arrow).
It’s also quite possible that Rust could underpin a product like Spark, with a scripting language on top.
The compilation time is a hurdle, but it’s not as if compiled languages are absent in numerical computing, for example Fortran.
Moreover, Logo-based and Lisp programming environments are more popular than other people think. Netlogo is heavily used in social sciences. Hard to think that even Python and R can displace those since they are so well suited for that kind of programming (agent manipulation).
I write Java and Rust at work, and i more or less agree.
More or less because there is some overlap: for example, people write low-latency trading systems in both. But it's very hard to get Java to perform well at the kind of work where Rust shines naturally.
There is also very little overlap between Rust and Go. There's far more between Java and Go than Rust and either!
20 years ago, almost none of that stuff existed. It's entirely possible that whatever exists 20 years from now will all be Rust. I don't think that's likely, but I think it's more likely than the landscape you describe still existing in any recognizable form.
Being as experienced as you are, couldn't you just learn Rust like any other language if you need it in your work (or just to try it out). Your post has a weird negative vibe, like it would cost you greatly to try it.
> As a big data engineer, Rust does not provide any advantage to me. Good luck rewritting all the big data tools from Java (Elasticsearch, Spark, Kafka, Hadoop, Neo4j, Deeplearning4j, Cassandra, Solr, Arrow, OrientDB)
Yep, this is true. Java has an upper hand in data engineering. There are better languages for targeting the Java libraries and JVM than Java itself. I always use Clojure instead of Java when doing data crunching because it is much easier to create parallel programs with it. Kotlin is also coming up in this space.
Having said that. I like Rust for some reasons:
- targeting multiple libc (musl, no libc at all)
- targeting multiple platforms (ppc, mips, etc., even though some of those are tierN)
- having ML features
Why I do not like Rust:
- accidental complexity and convoluted syntax (maybe it is only me, but it is very hard to follow what is happening)
For me the ideal language would be something like F# with the good parts of Rust.
Chrome is "on track"? Last I remember they had a document explaining why it's not a good idea for them. I disagree with it, and would very much wish for Chrome to ship Rust, but they don't seem to be interested for now.
They're referring to this[0] I believe, which is not glowing, but does seem mildly optimistic. Specifically:
> For now, Chrome investment in Rust will remain a background investigation (mostly directed towards prototyping these tools and techniques). If we become convinced this sort of interoperability is possible, we’ll revisit widespread use of Rust in Chrome
Yes that's the same document I refer to as well. So nothing has changed. A "mildly optimistic" rejection is a rejection of Rust nonetheless. Rust is definitely not "on track" of being adopted in Chromium. For that you'd need more. Right now it's rather a "we are interested, maybe we'll add it". Also note that Chromium engineers also built some languages themselves that solve some of the same problems that Rust solves: https://github.com/google/wuffs
Right, and their issues there all pertain to interoperability, not to issues with the Rust language and its implementation. They seem to view Rust as production-ready, and a language that would bring relevant benefits to the Chromium project, were it not for the C++ interoperability difficulties.
I think it's weird to blame Rust for that. I don't want to rush to blame Windows 7, but do consider that Windows 7 is massively EOL (2015; extended support also ended early this year) and I would not fault Mozilla for not doing any testing on that platform anymore. I don't think this has much, or anything, to do with Rust.
Rust is new.. the os software is older than the language itself and eol. Not surprised it wouldn't be tested or supported
Because of that I have less confidence rust will work better or improve my experience. So rewriting everything in rust worries me. It is an anti-selling point.
If a language is living in the now and targeting the future it might not be the best for those living in the past.
> I always hoped Java would make a comeback, but Rust is looking like the real deal.
I was mainly responding to this statement which came across quite aggressive. Java and C++ do not need comebacks and my comment was intended to be a wake-up call to reality. Yes, it is great to learn Rust on the side, but advocating something to be used everywhere is another story.
I don't think rewrites should be the primary talking point, but rather using it for new applications, especially in memory safety critical environments like kernel modules.
In my opinion data processing / ML / database development is one of the best use case for rust.
You want stable memory management and performance . This is still something than often times is done in C/C++.
There are several promising technologies being developer in rust in that space; tikv (key-value store), materialize (steaming DB), arrow/parquet has a rust implementation, Toshi (Elasticsearch), Tantivy (Lucene), etc
A good point and a long list of tools. But imo this is not the industry to say “good luck rewriting X” in. It is inevitable someone will come along to rewrite or write a new and improved tool for everything on that list. And it might be in Rust.
Stuff written before Rust existed was not written in Rust? I’m shocked. That’s like going back to the 90’s and pointing to all the software not written in Java.
And as for browsers, you know Mozilla created Rust specifically for Firefox?
I am a major fan of Rust. I upvoted your comment because it is mostly factual, and correct!
I agree that Rust is still early for most adopters.
There are a subset of people who need to solve problems that would normally be solved in C or C++, but they don't know or don't want to use those languages. Those people generally love Rust (I am in this camp).
Then there is everyone else, who either like Rust because it's a cool language, or don't for any number of reasons all of which are fine!
We are heading towards distributed high available SQL and NoSQL databases (with Redis, Postgres and MySQL compatibility for easier migration) designed to work exceptionally well in cloud environments like CockroachDB (Go), TiKV (Rust), Vitess (Go) to name a few.
That is because SRE teams want to be able to scale horizontally and make a node, rack or even region failure to be a normal thing, their company don't need to worry about.
The part you seem to be replying to doesn't imply that Rust is or should replace Java, at least not to me.
> I always hoped Java would make a comeback, but Rust is looking like the real deal.
To me it seems like SoSoRoCoCo is hoping their preferred language would make a comeback in popularity while also acknowledging that Rust is gaining popularity and real-world demand.
Why do you need to rewrite them? Rust can call C++ code, just like Java and Python can. (I note you list Python libraries like Tensorflow and PyTorch as C++, which is accurate because they're written in C++, but they're certainly heavily used from a non-C++ language.)
Rust definitely makes sense anywhere you are currently using C/C++ or Python/Ruby.
Java is somewhat of a special case though, especially the Big Data ecosystem. Much of the fastest and most important Java tools are written in pure Java (or other JVM language) and as such can't be called from Rust in a reasonable way.
Writing Big Data code in Rust is certainly possible from a performance perspective but is an uphill battle of lack of libraries, integration, etc. Some of this might change in time but I don't see Rust ever displacing Java from Big Data.
One of the big reasons for this is the main downside of Java is it's GC, which is mainly problematic for latency sensitive applications. In the Big Data field this is substantially reduced as a problem as the code is usually throughput bound and executed outside of interactive contexts, hiding the cost of the GC where it might be more apparent in other applications.
I love Rust and I think it has a lot of potential to replace some of the databases and networked services used in data infrastructure, just don't see it displacing Java as the top dog.
You've got it backwards. Rust is lower level than Java, so Java would be calling Rust, not the other way around.
Rust isn't aiming to displace Java. They're two different levels of abstraction. It's the same with Python. Rust isn't aiming to displace Python just as Java isn't aiming to displace Python either.
Rust one day could displace C++, but the odds are low, because in response to Rust C++ has stepped up its game and has changed as a language drastically. Most modern C++ features are concepts from Rust.
It's easy to call C++ code from Rust but it's often not trivial to wrap a C++ library with a safe Rust interface. C++ APIs often have implicit ownership structures that are difficult to encode using Rust's lifetimes.
To be fair, that doesn't put Rust in a worse position than any other language binding a C++ API. It's just that Rustaceans tend to have higher expectations in terms of safety.
> As a big data engineer, Rust does not provide any advantage to me
I assume that this is meant to imply that Rust doesn't provide advantages in the data engineering space. So I'll add in here -- as a data engineer myself -- that Rust is definitely providing an advantage for me. Combining it with Python (using PyO3 bindings) lets me write performant, safe code that interops with Python in a project for which Scala+Spark isn't feasible. So while Rust may not provide an advantage for you, that doesn't mean it doesn't have a potential niche for data engineering.
Cassandra has already been ported to C++ with Scylla. We switched to Scylla a year or so ago, haven't looked back.
As for the others, if it can be shown that porting them comes with a decrease in latency and infrastructure costs then they'll be ported. I've often wondered why they haven't been already.
