[blindseer]

I'm excited to see `--trim` finally make it, but it only works when all code from entrypoints are statically inferrable. In any non-toy Julia program that's not going to be the case. Julia sorely needs a static mode and a static analyzer that can check for correctness. It also needs better sum type support and better error messages (static and runtime).

In 2020, I thought Julia would be _the_ language to use in 2025. Today I think that won't happen until 2030, if even then. The community is growing too slowly, core packages have extremely few maintainers, and Python and Rust are sucking the air out of the room. This talk at JuliaCon was a good summary of how developers using Rust are so much more productive in Rust than in Julia that they switched away from Julia:

https://www.youtube.com/watch?v=gspuMS1hSQo

Which is pretty telling. It takes a overcoming a certain inertia to move from any language.

Given all that, outside of depending heavily on DifferentialEquations.jl, I don't know why someone would pick Julia over Python + Rust.

[wrathofmonads]

I don't think Julia was designed for pure overhead projects in memory-constrained environments, or for squeezing out that last 2% of hardware performance to cut costs, like C++, Rust or Zig.

Julia is the language to use in 2025 if what you’re looking for is a JIT-compiled, multiple-dispatch language that lets you write high-performance technical computing code to run on a cluster or on your laptop for quick experimentation, while also being metaprogrammable and highly interactive, whether for modelling, simulation, optimisation, image processing etc.

[throwawaymaths]

actually I think it sort of was, I remember berkeley squeezing a ton of perf out of their cray for a crazy task because it was easy to specialize some wild semi-sparse matrix computations onto an architecture with strange memory/cache bottlenecks, while being guaranteed that the results are still okay.

[leephillips]

Yes. Remember that the four languages that have been used for very high performance computing are Fortran, C, C++, and Julia:

https://www.hpcwire.com/off-the-wire/julia-joins-petaflop-cl...

For large-scale physics simulations, Fortran or Julia are the obvious choices.

[tagrun]

Telling what? Did you actually listen to the talk that you linked to, or read the top comment there by Chris Rackauckas?

> Given all that, outside of depending heavily on DifferentialEquations.jl, I don't know why someone would pick Julia over Python + Rust.

See his last slide. And no, they didn't replace their Julia use in its entirety with Rust, despite his organization being a Rust shop. Considering Rust as a replacement for Julia makes as much sense to me as to considering C as a replacement for Mathematica; Julia and Mathematica are domain specific (scientific computation) languages, not general systems programming languages.

Neither Julia nor Mathematica is a good fit for embedded device programming.

I also find it amusing how you criticize Julia while praising Python (which was originally a "toy" scripting language succeeding ABC, but found some accidental "gaps" to fit in historically) within the narrative that you built.

> In any non-toy Julia program that's not going to be the case.

Why?

[blindseer]

> Telling what? Did you actually listen to the talk that you linked to, or read the top comment there by Chris Rackauckas?

To clarify exactly where I'm coming from, I'm going to expand on my thoughts here.

What is Julia's central conceit? It aims to solve "the two language" problem, i.e. the problem where prototyping or rapid development is done in a dynamic and interactive language like Python or MATLAB, and then moved for production to a faster and less flexible language like Rust or C++.

This is exactly what the speaker in the talk addresses. They are still using Julia for prototyping, but their production use of Julia was replaced with Rust. I've heard several more anecdotal stories of the exact same thing occurring. Here's another high profile instance of Julia not making it to production:

https://discourse.julialang.org/t/julia-used-to-prototype-wh...

Julia is failing at its core conceit.

Julia as a community have to start thinking about what makes a language successful in production.

Quote from the talk:

> "(developers) really love writing Rust ... and I get where they are coming from, especially around the tooling."

Julia's tooling is ... just not good. Try working several hundred thousand line project in Julia and it is painful for so many reasons.

If you don't have a REPL open all the time with the state of your program loaded in the REPL and in your head, Julia becomes painful to work in. The language server crashes all the time, completion is slow, linting has so many false positives, TDD is barebones etc. It's far too easy to write type unstable code. And the worst part is you can write code that you think is type stable, but with a minor refactor your performance can just completely tank. Optimizing for maintaining Julia code over a long period of time with a team just feels futile.

That said, is Python perfect? Absolutely not. There's so many things I wish were different.

But Python was designed (or at the very least evolved) to be a glue language. Being able to write user friendly interfaces to performant C or C++ code was the reason the language took off the way it did.

And the Python language keeps evolving to make it easier to write correct Python code. Type hinting is awesome and Python has much better error messages (static and runtime). I'm far more productive prototyping in Python, even if executing code is slower. When I want to make it fast, it is almost trivial to use PyO3 with Rust to make what I want to run fast. Rust is starting to build up packages used for scientific computing. There's also Numba and Cython, which are pretty awesome and have saved me in a pickle.

As a glue language Python is amazing. And jumping into a million line project still feels practical (Julia's `include` feature alone would prevent this from being tenable). The community is growing still, and projects like `uv` and `ty` are only going to make Python proliferate more.

I do think Julia is ideal for an individual researcher, where one person can keep every line of code in their head and for code that is written to be thrown away. But I'm certainly not betting the near future on this language.

[wolvesechoes]

Python has useful and rich ecosystem that grows every day. Julia is mostly pile of broken promises (it neither reads as Python, nor it runs as C, at least not without significant effort required to produce curated benchmarks) and desperate hype generators.

[tagrun]

Since you have a rosy picture of Python, I assume you're young. Python has been mostly a fringe/toy language for 2 decades, until around ~2010, when a Python fad started not too different from the Rust fad of today, and at some point Google started using it seriously and thought they can fix Python but gave up eventually. The fad lived on and kept evolving and somehow found some popularity with SciPy and then ML. I used it in 90s a little, and I found the language bad for anything other than replacing simple bash scripts or simple desktop applications or a desktop calculator, and I still think it is (but sure, there are people who disagree and think it is a good language). It was slow and didn't have type system, you didn't know whether your code would crash or not until you run that line of code, and the correctness of your program depended on invisible characters.

"Ecosystem" is not a part of the language, and in any case, the Python ecosystem is not written in Python, because Python is not a suitable language for scientific computing, which is unsurprising because that's not what it was designed for.

It is ironic you bring up hype to criticize Julia while praising Python which found popularity thanks to hype rather than technical merit.

What promise are you referring to? Who promised you what? It's a programming language.

[wolvesechoes]

> Ecosystem" is not a part of the language, and in any case, the Python ecosystem is not written in Python, because Python is not a suitable language for scientific computing

Doesn't matter. Languages do not matter, ecosystems do, for they determine what is practically achievable.

And it doesn't matter that Python ecosystem relies on huge amounts of C/C++ code. Python people made the effort to wrap this code, document it and maintain those wrappers. Other people use such code through Python APIs. Yes, every language with FFI can do the same. For some reason none achieved that.

Even people using Julia use PythonCall.jl, that's how much Python is unsuitable.

> What promise are you referring to? Who promised you what? It's a programming language.

Acting dumb is poor rhetorical strategy, and ignores such a nice rhetorical advice as principle of charity - it is quite obvious that I didn't mean that programming language made any promise. Making a promise is something that only people can do. And Julia creators and people promoting it made quite bombastic claims throughout the years that turned out to not have much support in reality.

I leave your assumptions about my age or other properties to you.

[wrathofmonads]

Ecosystems matter, but runtimes do as well. Take Java, for instance. It didn’t have to wrap C/C++ libraries, yet it became synonymous with anything data-intensive. From Apache Hadoop to Flink, from Kafka to Pulsar. Sure, this is mostly ETL, streaming, and databases rather than numeric or scientific computing, but it shows that a language plus a strong ecosystem can drive a movement.

This is why I see Julia as the Java for technical computing. It’s tackling a domain that’s more numeric and math-heavy, not your average data pipeline, and while it hasn’t yet reached the same breadth as Python, the potential is there. Hopefully, over time, its ecosystem will blossom in the same way.

[tagrun]

If what determines the value of a language libraries (which makes no sense to me at all, but let's play your game), then it is one more argument against Python. You don't need FFI to use a Fortran library from Fortran, and I (and many physicists) have found Fortran better suited to HPC than Python since... the day Python came to existence. And no, many other scripting languages have wrappers, and no, scientific computing is not restricted to ML which the only area Python can be argued to have most wrapper libraries to external code.

Language matters, and two-language problem is a real problem, and you can't make it go away by closing your ears and chanting "doesn't matter! doesn't matter!"

Julia is a real step toward solving this problem, and allows you to interact with libraries/packages in ways that is not possible in Python + Fortran + C/C++ + others. You are free to keep pretending that problem doesn't exist.

You are making disparaging and hyperbolic claims about hyperbolic claims without proper attribution, and when asked for source, you cry foul and sadly try to appear smart by saying "you're acting dumb". You should take on your advice and instead of "acting dumb", explicitly cite what "promises" or "bombastic claims" you are referring to. This is what I asked you to do, but instead of doing it, you are doing what you are doing, which is interesting.

[wolvesechoes]

> If what determines the value of a language libraries (which makes no sense to me at all, but let's play your game), then it is one more argument against Python

The fact that you can use those nice numerical and scientific libraries from the language that had also tremendous amount of nice libraries from other domains, wide and good IDE support, is very well documented and has countless tutorials and books available... is an argument against that language? Because you can easily use Fortran code in Fortran?

Nice.

> You don't need FFI to use a Fortran library from Fortran

Wow. Didn't know that.

> And no, many other scripting languages have wrappers,

Always less complete, less documented, with less teaching materials available etc.

But sure, many other languages have wrappers. Julia for example wraps Python API.

> and no, scientific computing is not restricted to ML

Never said it is. I don't do ML, by the way.

> You are making disparaging and hyperbolic claims about hyperbolic claims without proper attribution, and when asked for source, you cry foul

Yeah, yeah. My claims on marketing like "Julia writes like Python, runs like C" are hyperbolic and require explicit citation, even though everyone that had any exposure to this language knows such and similar catch-phrases.

Look, you like Julia, good for you. Have fun with it.

[throwawaymaths]

in the early aughts educators loved the shit out of python because "it forced kids to organize their code with indentation". This was about a decade before formatting linters became low key required for languages.

[mccoyb]

These are exactly the feelings that I left with from the community in ~2021 (along with the AD story, which never really materialized _within_ Julia - Enzyme had to come from outside Julia to “save it” - or materialized in a way (Zygote) whose compilation times were absolutely unacceptable compared to competitors like JAX)

More and more over time, I’ve begun to think that the method JIT architecture is a mistake, that subtyping is a mistake.

Subtyping makes abundant sense when paired with multiple dispatch — so perhaps my qualms are not precise there … but it also seems like several designs for static interfaces have sort of bounced off the type system. Not sure, and can’t defend my claims very well.

Julia has much right, but a few things feel wrong in ways that spiral up to the limitations in features like this one.

Anyways, excited to check back next year to see myself proven wrong.

[jakobnissen]

I basically agree with subtyping (but not multiple dispatch). More importantly, I think it's important to recognize that Julia has a niche that literally no one can compete with - interactive, dynamic and high performance.

Like, what exactly is the alternative? Python? Too slow. Static languages? Unusable for interactive exploration and data science.

That leaves you with hybrids, like Python/Cython, or Python/Rust or Numba, but taken on their own term, these are absolutely terrible languages. Python/Rust is not safe (due to FFI), certainly not pleasant to develop in, and no matter how you cut your code between the languages, you always lose. You always want your Python part to be in Rust so you get static analysis, safety and speed. You always want your Rust part to be in Python, so you can experiment with it easier and introspect.

[mccoyb]

To clarify my comment: I agree that multiple dispatch is a very good language feature. I enjoy it, and I’m well-versed in the expression problem, yada yada.

That’s not what I meant by “method JIT architecture” — I meant calling back into the compiler at runtime to specialize code when the types are known.

[blindseer]

I think multiple dispatch (useful as it is) is a little overrated. There's a significant portion of the time where I know I have a closed set of cases to cover, and an enum type with a match-like syntax would have worked better for that. For interfaces, multiple dispatch is good but again I would have preferred a trait based approach with static type checking.

I largely think multiple dispatch works well in Julia, and it enables writing performant code in an elegant manner. I mostly have smaller gripes about subtyping and the patterns it encourages with multiple dispatch in Julia, and larger gripes about the lack of tooling in Julia.

But multiple dispatch is also a hammer where every problem in Julia looks like a nail. And there isn't enough discussion, official or community driven, that expands on this. In my experience the average developer to Julia tends to reach for multiple dispatch without understanding why, mostly because people keep saying it is the best thing since sliced bread.

wrt to hybrid languages, honestly, I think Python/Cython is extremely underrated. Sure you can design an entirely new language like Mojo or Julia, but imo it offers only incremental value over Python/Cython. I would love to peek into another universe where all that money, time and effort for Mojo and Julia went to Cython instead.

And I personally don't think Python/Rust is as bad. With a little discipline (and some tests), you can ensure your boundary is safe, for you and your team. Rust offers so much value that I would take on the pain of going through FFI. PyO3 simplifies this significantly. The development of `polars` is a good case study for how Rust empowers Python.

I think the Julia community could use some reflection on why it hasn't produced the next `polars`. My personal experience with Julia developers (both in-person and online) is that they often believe multiple dispatch is so compelling that any person that "saw the light" would obviously naturally flock to Julia. Instead, I think the real challenge is meeting users where they are and addressing their needs directly. The fastest way to grow Julia as a language is to tag along Python's success.

Would I prefer a single language that solves all my problems? Yes. But that single language is not Julia, yet, for me.

PS: I really enjoy your blog posts and comments.

[wrathofmonads]

Mojo has a different scope than Julia and Python, it targets inference workloads.

Polars is a dataframe library. Yes, it features vectorized operations, but it is focused on columnar data manipulation, not numerical algorithm development. I might say that this is narrow framing, people are looking at Julia through the lens of a data scientist and not of an engineer or computational scientist.

[blindseer]

Most of my gripes are when trying to use Julia the way a software engineer would use a programming language.

Most "data scientist" code is exploratory (it's a prototype or a script for an one-off exploration) in nature. And my main gripe is that making that code production ready and maintainable over a long period of time is so difficult that I would switch to Rust instead. If I were going to switch to Rust, I might as well start with Python.

[pjmlp]

There is no "_the_ language to use", I always pick the language based on the project delivery requirements, like what is tier 1 support on a specific SDK, not pick the project based on the language.