Julia for Matlab users | Hacker News Reader

Julia for Matlab users

sveme.org

59 pointsluu23 comments

I've been trying out Julia for a couple weeks. I love the language, it's simple, powerful, and fast. My problem is that I've run into a lot of cases where I have an error, but no indication of where the error is. Sometimes it's a syntax error somewhere in the file, other times it's a stack trace that gives me my top-level function call and a line somewhere in the standard library, but nothing in between.

So I'm finding I can write the code very quickly, but if I get an error like this I lose all the time I gained and more. I'm sure this is a matter of it being such a young language, and also that someone more experienced would avoid many of the pitfalls catching me. But for noobs like me it's difficult. I hope they improve this soon.

I've been playing around with Julia for about a week now too. There are so many things to like, but I agree - the tracebacks are pretty useless half of the time.

The error messages are driving me nuts as well. Otherwise, Julia is a very impressive language that can do more than replace Matlab.

The only other irk I have is that an ASCIIString doesn't have UTF8String as a parent type (ASCIIString <: UTF8String -> false)

Which version of Julia are you using and on what platform? Windows backtraces are pretty bad, but other platforms decent. The situation on Windows is likely to improve soon, fortunately. If you have specific problems that are reproducible, it would be great to have bug reports so we can improve things: https://github.com/JuliaLang/julia/issues.

Version 0.2.0 on Ubuntu.

I should have kept better notes. Going forward, I'll try and reproduce in small amounts of code and file reports.

One was pretty silly, I used "end" as a variable name. It blew up with a syntax error on a line where I was assigning to it, but didn't tell me what line or what variable name. (But it did say something about an unexpected assignment, so it knew where the problem was, it just didn't tell me.)

I've got the stacktrace issue right now with an error regarding an incorrect byte index into a unicode string. I get why that happens, just can't find what's causing it in my code.

Edit: changed to an ascii string after seeing someone mention it here, now I'm getting a BoundsError in copy! at array.jl:49, and again the stacktrace tells me nothing else besides the top-level function call.

Matlab's strength does not only lie in the simplicity of it's code or in not having to load libraries explicitly. It also lies in the fairly good IDE. Built for scientists who want to prototype code quickly, there is an interpreter and an editor right there for you. Not to mention that you never have to configure anything, it just works.

I love Julia, but for wider adoption, it will definitely need integration with a good IDE that "hides complexity from it's users".

There is Julia Studio[0], but I'm not sure it is competitive as of yet.

[0]: http://forio.com/products/julia-studio/

I've been looking at this, but it doesn't yet have watch windows and package import is somewhat broken. That said I do like the interface.

You can use Julia in IPython, but I'm not sure if that gets you everything you'd want. It definitely doesn't meet the "no configuration" criterion.

This is a big issue for some people in my lab too. There are interesting efforts towards this, check out Julietta:

https://github.com/tknopp/Julietta.jl

Though not really an IDE, IJulia is kind of neat, reminds me of mathematica in some ways (like IPython).

Julia intrigues me and I'm on their mailing list but daily perusal of activity on that list has shown me clearly that Julia is still a long way from deciding what it will be when it grows up.

I quite like it and look forward to all the conceptual issues being settled and a stable version delivered. Lotsa' smart people providing input on that but that's both a blessing and a curse.

To much variety of intrinsic data aggregates for me but they justify that on it being compiled and very performance sensitive. Thus a lot of "you can do x on an A but not y. But if you must y, a B is provided instead.'

I'm not very familiar with Julia, could you provide an example of what you're taking about in the last paragraph?

https://groups.google.com/forum/#!topic/julia-users/7DXD7Qbk...

Starting with Stefan Karpinski's first comment.

So I want to plot a few lines and label my axes. First step? Install 2 of the 3, ... er, 4 available plotting packages, named "Winston" and "Gertrude," or "Jenkins" and "Marsalis," or something else equally meaningless. Once I've done this, I can issue a bunch of low-level commands to produce my simple plot.

Matlab may be an ugly language, but offering this kind of amateur software as a replacement is just pathetic. Octave, R, and various python packages have been viable for years.

Assuming you have matplotlib installed:

    using PyPlot
    x = linspace(0,2*pi,1000); y = sin(3*x + 4*cos(2*x));
    plot(x, y, color="red", linewidth=2.0, linestyle="--")
    title("A sinusoidally modulated sinusoid")

You can use Gadfly to plot using the grammar of graphics (as implemented by R's ggplot2), which is the nicest plotting API I've seen:

http://dcjones.github.io/Gadfly.jl/

Julia aims for a different niche than R or Python, and is far more performant than Octave.

> Julia aims for a different niche than R or Python, and is far more performant than Octave.

Julia's loops are more performant. Its vectorised code is not.

http://stackoverflow.com/questions/19208014/is-the-julia-lan...

It is open-source software, not a walled garden - there are going to be multiple options, and they are not going to all be named ["PLOTTINGISHERE$i" for i in 1:4].

If you like MATLAB/pyplot/matplotlib style, PyPlot.jl is for you.

If you like ggplot2, Gadfly.jl is for you.

If you like gnuplot, Gaston.jl is for you.

Finally, but not least, is Winston.jl, which is its own thing altogether.

I think in many cases its good for packages that overlap to come together/one to become dominant, but I don't think that applies here.

https://github.com/stevengj/PyPlot.jl

https://github.com/dcjones/Gadfly.jl

https://github.com/mbaz/Gaston.jl

https://github.com/nolta/Winston.jl

I guess Gaston would be my choice, since I'm most familiar with Gnuplot syntax for tweaking plots. Still, it's a shame Julia inherited the Ruby/Rails convention of making up nonsense names for packages. While they don't need to be named PLOTTING$i, it wouldn't hurt for them to be named Plot::Gnuplot, etc. (or whatever the syntax is for nested names). Doing otherwise is just putting branding ahead of actually trying to help people.

For those new to Julia, you can also try it out here: http://forio.com/julia/repl/

(Full disclosure, I'm affiliated)

CSV is rarely used with Matlab. Can Julia read HDF, H5, or CDF formats?

Yes, there are packages that can read and write HDF5 (including specific support for .mat) [0] and NetCDF [1] formats.

[0]: https://github.com/timholy/HDF5.jl [1]: https://github.com/meggart/NetCDF.jl

Reply on news.ycombinator.com