4 ms·

If anyone is looking for an alternative to R or python, there's Julia already.

CameronNemo 6y ago
Also Rust has a good datagrams library now, polars. Not as mature an ecosystem as Julia, but hopefully it improves in the future.
- hu3 6y ago
  I find Rust's borrow checker too clunky for exploratory work. It breaks my flow and imposes higher cognitive load. The slow compiler doesn't help either.
  - Gibbon1 6y ago
    I tend to think the innately slow compiler is basically a fatal mistake. Rust will never be able to be used for large projects. It's not so obvious now because everything it's used for is tiny.
    - CameronNemo 6y ago
      Hopefully some of the work out of cranelift, gccrs, and/or rust-analyzer can be used to speed up compilation.
      - Gibbon1 6y ago
        Seriously one hopes. I think of projects in C++ that takes half an hour to compile, in rust they would take half a day or longer.
        c-cube 6y ago
        What supports that statement exactly? If you split your big projects into crates it would recompile pretty fast. C++ can also take ages to compile (eg. compiling Firefox from scratch). Keeping your code modular to get decent compile times seems like a win win.
        pjmlp 6y ago
        I never compile C++ projects from scratch like I am forced to do with Rust. The only code I compile from scratch in C++ is the code I write myself, everything else is available as binary libraries, something that cargo doesn't do, and it is not part of the near future roadmap, if ever. Then, after compiled, most of the stuff lands on the VC++ metadata files, so incremental compilation and linking cuts even more time from the usual edit-compile-debug workflow.
        c-cube 6y ago
        Interesting, is this common on windows? On Linux I've never seen precompiled C++ libraries (at least not these with templates) back when I compiled stuff more often (read: back when I used gentoo). Do g++ and clang++ support precompiled libraries in the general case? I suppose C++ modules might make it more common anyway, but I don't see why rust couldn't do it if they ever prioritize it.
        lanstin 6y ago
        The distro takes care of it and you just yum / apt-get / whatever the lib and then compile the code you typed in. Template libs will slow down your compile times but there is still a lib boost.so etc sitting around.
  - CameronNemo 6y ago
    That is fair. I have not done enough exploratory work in Rust to comment. Maybe there are abalysis patterns that can avoid bumping up against the borrow checker.
  - mountainriver 6y ago
    Yeah Rust is really close to what I want but I agree the borrow checker adds a bit too much overhead when dealing with data science. It seems like reference counting is probably the move here
sundarurfriend 6y ago
If this project can maintain Go's fast compile times and ability to make reliable, concice binaries, those would be two big pluses in areas where Julia is currently weak. That would make this a good choice in projects where those are high priorities.
mountainriver 6y ago
I think what people want is all the great things Go brings to the table but just geared a little more towards data work. Julia offers a lot in the data world and not much in the engineering world.

Also Rust has a good datagrams library now, polars. Not as mature an ecosystem as Julia, but hopefully it improves in the future.

hu3 6y ago
I find Rust's borrow checker too clunky for exploratory work. It breaks my flow and imposes higher cognitive load. The slow compiler doesn't help either.
- Gibbon1 6y ago
  I tend to think the innately slow compiler is basically a fatal mistake. Rust will never be able to be used for large projects. It's not so obvious now because everything it's used for is tiny.
  - CameronNemo 6y ago
    Hopefully some of the work out of cranelift, gccrs, and/or rust-analyzer can be used to speed up compilation.
    - Gibbon1 6y ago
      Seriously one hopes. I think of projects in C++ that takes half an hour to compile, in rust they would take half a day or longer.
      - c-cube 6y ago
        What supports that statement exactly? If you split your big projects into crates it would recompile pretty fast. C++ can also take ages to compile (eg. compiling Firefox from scratch). Keeping your code modular to get decent compile times seems like a win win.
        pjmlp 6y ago
        I never compile C++ projects from scratch like I am forced to do with Rust. The only code I compile from scratch in C++ is the code I write myself, everything else is available as binary libraries, something that cargo doesn't do, and it is not part of the near future roadmap, if ever. Then, after compiled, most of the stuff lands on the VC++ metadata files, so incremental compilation and linking cuts even more time from the usual edit-compile-debug workflow.
        c-cube 6y ago
        Interesting, is this common on windows? On Linux I've never seen precompiled C++ libraries (at least not these with templates) back when I compiled stuff more often (read: back when I used gentoo). Do g++ and clang++ support precompiled libraries in the general case? I suppose C++ modules might make it more common anyway, but I don't see why rust couldn't do it if they ever prioritize it.
        lanstin 6y ago
        The distro takes care of it and you just yum / apt-get / whatever the lib and then compile the code you typed in. Template libs will slow down your compile times but there is still a lib boost.so etc sitting around.
- CameronNemo 6y ago
  That is fair. I have not done enough exploratory work in Rust to comment. Maybe there are abalysis patterns that can avoid bumping up against the borrow checker.
- mountainriver 6y ago
  Yeah Rust is really close to what I want but I agree the borrow checker adds a bit too much overhead when dealing with data science. It seems like reference counting is probably the move here

sundarurfriend 6y ago

If this project can maintain Go's fast compile times and ability to make reliable, concice binaries, those would be two big pluses in areas where Julia is currently weak. That would make this a good choice in projects where those are high priorities.

mountainriver 6y ago

I think what people want is all the great things Go brings to the table but just geared a little more towards data work. Julia offers a lot in the data world and not much in the engineering world.