Getting started with the Julia language

Tutorials & Resources
1

The Julia language is not a general purpose programming language , like C or Rust. It is specialised for numerical calculations of the type commonly used by mathematicians or scientists. As such it is an alternative to Matlab , Octave or maybe R.
The feature of Julia that interests me is that it makes parallel programming easy… ie I can write code that will use all available CPU threads. That helps with large calculations.

I had one previous try at learning Julia, but it stalled. I am making a fresh start here.

Installing Julia

The official JUlia website is here

It offers various types of download and it pushes very strongly for one to use the downloadable version of Julia, and not some earlier version from your package system. That goes a bit against my grain, but I did it,

trinity:[nevj]:~$ cat wget.out
trinity:[nevj]:~$ wget https://julialang-s3.julialang.org/bin/linux/x64/1.11/julia-1.11.4-linux-x86_64.tar.gz
--2025-03-28 20:26:15--  https://julialang-s3.julialang.org/bin/linux/x64/1.11/julia-1.11.4-linux-x86_64.tar.gz
Loaded CA certificate '/etc/ssl/certs/ca-certificates.crt'
Resolving julialang-s3.julialang.org (julialang-s3.julialang.org)... 2a04:4e42:62::561, 199.232.138.49
Connecting to julialang-s3.julialang.org (julialang-s3.julialang.org)|2a04:4e42:62::561|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 266267808 (254M) [application/x-tar]
Saving to: ‘julia-1.11.4-linux-x86_64.tar.gz’

julia-1.11.4-linux- 100%[===================>] 253.93M  16.6MB/s    in 17s     

2025-03-28 20:26:48 (14.8 MB/s) - ‘julia-1.11.4-linux-x86_64.tar.gz’ saved [266267808/266267808]julia-1.11.4

I did the install as ‘nevj’ , not root.
It was very fast. What I ended up with was a file called julia-1.11.4 in my home directory

trinity:[nevj]:~/julia-1.11.4$ ls -F
bin/  etc/  include/  lib/  libexec/  LICENSE.md  share/

The whole install is there inside that file, and nothing else is written anywhere on your system. You can move it to anywhere you want, and to uninstall you just remove the file recursively.

NOTE: I have since discovered that the above wget download is unnecessary.
One only needs to do thhe curl download ( see below)
I have since deleted the ~/julia-1.11.4 directory.
I am not sure what it is for? Juliaup does not use it.

So far, so good. Now can I run it?

trinity:[nevj]:~$ which julia
nothing found

You have to put the location of the Julia binary in your PATH.

export PATH=$PATH:/common/bin:/home/nevj/julia-1.11.4/bin

Then the julia binary can be found

trinity:[nevj]:~/juliawork$ julia
               _
   _       _ _(_)_     |  Documentation: https://docs.julialang.org
  (_)     | (_) (_)    |
   _ _   _| |_  __ _   |  Type "?" for help, "]?" for Pkg help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 1.11.4 (2025-03-10)
 _/ |\__'_|_|_|\__'_|  |  Official https://julialang.org/ release
|__/                   |

julia>

You get a prompt julia>. That interface is called REPL… Read, Evaluate, Print Loop. You can test things interactively there, for example

julia> function add(a,b)
       sum = a+b
       return sum
       end
add (generic function with 1 method)

julia> add(3,4)
7

julia> exit()

OK, I tested a simple function… it works.

Updating Julia

I then discovered that I had missed out on installing the ‘juliaup’ utility which looks after updates.
So I went back and did

curl -fsSL https://install.julialang.org | sh

and this time it made directories .julia and .juliaup, and it left my julia-1.11.4 directory alone. It also added more lines to my .bashrc as follows

# >>> juliaup initialize >>>

# !! Contents within this block are managed by juliaup !!

case ":$PATH:" in
    *:/home/nevj/.juliaup/bin:*)
        ;;

    *)
        export PATH=/home/nevj/.juliaup/bin${PATH:+:${PATH}}
        ;;
esac

# <<< juliaup initialize <<<

and I discover there is a startup file in /home/nevj/.juliaup/bin which finds the binary automatically… so I dont need the export statement I just added.

I am not sure about this complicated double stage install, and the Julia webpage does not explain it thoroughly, but I got there

Launching a Julia program from a program file

You dont have to work in the interactive REPL interface. You can put a Julia function in a text file with a .jl extension, and run it as follows

Make a file add2.jl
#! /home/nevj/.juliaup/bin/julia
function add(a,b)
 sum = a + b
 return sum
end
print(PROGRAM_FILE)
println(ARGS)
a = parse(Int64,ARGS[1])
b = parse(Int64,ARGS[2])
sum = add(a,b)
println(sum)

Note this file contains a function and a main program
Make it executable with chmod 755 add2.jl, then

trinity:[nevj]:~/juliawork$ julia add2.jl 4 5
add2.jl["4", "5"]
9
trinity:[nevj]:~/juliawork$

or

trinity:[nevj]:~/juliawork$ ./add2.jl 7 8
./add2.jl["7", "8"]
15

You can launch it either using julia, or as a script ( because I put in the #!.. line)

So it can handle CLI arguments, which is an essential tool.

Where to?

Big learning curve now. It took a week to get to there.
The base language is easy enough.
I need to find all the libraries of useful functions .

Some reactions

The REPL interface is a poor substitute for the equivalent R interface.
Saving work in there is cumbersome. Maybe its is my lack of practice.

There is something strange about library functions that I need to grasp.
Julia seems to be able to use them directly from the internet without keeping a local library? That is different to any language I have used.

6 Likes
2

Please forgive a dumb question: what is a practical application for Julia?

3 Likes
3

Hi Neville, :waving_hand:

thank you very much for the introduction to julia.

It clearly shows that you know a lot about programming languages in general and even more about specific ones. I´m impressed. :+1:

I myself have just a basic understanding of bash scripting. That´s all.
Although I wrote some scripts in the past it seems to be nothing compared to what you are able to achieve.

Congratulations on that, Neville. :clap:

Many greetings from Rosika :slightly_smiling_face:

4 Likes
4

But interested in how you plan to use it Neville…
I did lots of statistiques but excel did all i needed.

2 Likes
5

I thought it looked like juliaup is all you need to install. So, I went through the motions on my laptop and indeed that is the case. It installed the juliaup program and also the Julia executable itself.

It would seem the first step of manually installing Julia is not needed. You could create a link from the location you intended to install to the location juliaup installed it to and you’d be all set. Maybe just eliminate that first location and change to your path.

image
image1025×823 87.5 KB

I played with Julia a bit last year (I think). It would seem to be better suited to statistical applications than Python. The syntax looks pretty similar to me. The ecosystem for Python is quite a bit bigger and with freely available libraries you gain back any speed difference. These libraries are written in C or Rust or another lower level language.

Julia looks like pretty good stuff though. It very well could take better advantage of muli-processing or multi-threading than Python or make it easier.

2 Likes
6

You are right. I found out the hard way.

It certainly does it better than R.
R probably has the best of all statistical libraries, and many of the libraries, like BLAS and Lapack are setup to multiprocess, but R itself does not multiprocess. There are addons, but they are cumbersome.
I will write something on multiprocessing in Julia when I get there. Have only read about it so far.

There are some interesting operators, like

  • 1//3 the exact value of a rational number
  • dot operators for vectors
[1, 2, 3] .^ 3

means cube each element of the vector

It will take some practice to really see if I should move from R.
There is this document

Written by Doug Bates, a noted contributor to R.c

3 Likes
7

If you want to program a custom statistical analysis of some data, the choice is

  • R
  • Python
  • SAS
  • Julia

You can use a general purpose langusge like Fortran or C or Pascal, but it will be a slog, because they lack special statistal libraries.

If you want to compute how to rotate the space station, or some other high level physics calculation, you need a language which makes tensor calculus easy and efficient.The choices are

  • Matlab
  • Julia

So Julia is really for elaborate calculations requiring parallel computing. It is not the only option for that… lots of scientists still use Fortran… but it is modern and behaves like an interpreter rather than a compiler. It actually uses a Jit compiler, but you dont see it, it feels like Python or R.

See also the link in @callpaul.eu 's reply

2 Likes
8

I have a large complicate R program that does statistical estimation for genetics data.
It is slow, and runs out of memory on large datasets.
I am thinking of moving it to Julia… not translating it, I can run an R program inside Julia. Then I might try and rewrite in Julia the parts of it that are inefficient in R
So I end up with my nice R interface, but some more efficient code.
What I can do about memory usage is a bigger problem.
Using Julia will not change that, I need to change the algorithm.
The whole idea might yet fizzle out with unexpected difficultues. I have to try. It will be educational.

Given my usual haphazard planning, I may and up discovering something else that can usefully be done in Julia. One has to have an open mind at all times.
If you dont look, you dont find anything.

4 Likes
9

Hi Rosika,
That is a good start Rosika.
You could go on to any language from there if you wished.
The easiest is probably Python.
Regards
Neville

3 Likes
10

Hi Neville, :waving_hand:

Thanks for the recommendation.

Actually I looked into it a few years back but I wasn´t really able to grasp the concept of importing modules.
Perhaps I didn´t use a teaching source which suited me at the time. :thinking:
So I let it go in the end. Such a shame.

Seems I have to look for a better teaching method/source.

Many greetings from Rosika :slightly_smiling_face:

1 Like
11

R is not all that difficult either… as a language it is fairly simple and has an easy workspace setup.
Pascal is easy too.

I would not recommend Julia for starters… it
has too many esoteric features.
I am struggling just to read the manual.

1 Like
12

@nevj :

Thanks, Neville, for your evaluation.

I might try a new start with python one day. :wink:

Cheers from Rosika :slightly_smiling_face:

2 Likes
13

Thanks for the explanation. I believe I’ll leave that area to you and Paul, while I focus on trying to conquer golf and understand women. Those simple tasks are more suited to my brain.

3 Likes
14

Go for it! The Python community is large and pretty friendly. All the AI help works pretty well with Python because there is such a large base of code to draw on or train the AI on.

You could ask Copilot or another AI to explain a short Python script with an import statement and maybe that would help.

I don’t think it’s possible to conquer golf or understand women. But it’s not the destination, it’s the journey.

2 Likes
15

@pdecker :

Thank you very much.

As I have next to no initial knowledge of python I never could make head or tail of any python scripts whenever I looked at them.
I guess learning python should be a worthwhile effort all the same.

Thanks again and cheers from Rosika :slightly_smiling_face:

2 Likes
16

I think the same applies to mastering Julia.
My learning technique is

  • read the Manual
  • dive straight into a modest sized task … I am going to try and do some image processing for starters
  • dont ever hope to know it all

Did you notice…I put a Note in the first reply, retracting the wget step. It does seem completely unnecessary. I deleted the directory it made and Julia still runs?
I wonder if they could fix their misleading home page?

3 Likes
17

“I regard golf as an expensive way of playing marbles.”
GK Chesterton

4 Likes
18

You did not have enough data to process :slight_smile:
(Neither me :slight_smile: )
I see my son working with R, on really huuuge datasets in molecular biology.
I don’t understand a single word from him :smiley:
All I can do is to make R (and R-studio) to start on his computer.

1 Like
19

Not sure how much now, but excel did not have enough rows to import in one go, had to run it over 3 files and then combine the results, which gave slightly out readings on some things.

It was all about patient records in a hospital over a year, lengths of stay, admission source, final outcomes … not that interesting

4 Likes
20

That is great. Your son is in the right place for that work.

2 Likes