tags: - Free

Julia Language

Julia language is a high-performance, dynamic programming language. Julia is excellent for scientific computing because it can compile efficient native code using LLVM and includes mathematical functions, parallel computing capabilities, and a package manager in the standard library. Furthermore, Julia's syntax is intuitive and easy to learn, the multiple-dispatch paradigm allows writing composable code, increasing the ability to reuse existing code, and environments enable executing code in a reproducible way.

Julia Language

License

Julia language is licensed under free and open source MIT license.

Available

Julia language is available on Puhti, Mahti, and LUMI using the module system. On Puhti and Mahti, the Julia module is included on the module path by default. On LUMI, we must add the module files under CSC's local directory to the module path as follows.

module use /appl/local/csc/modulefiles

We can check the available versions as follows.

module avail julia

Usage

Loading the Julia module

We can load the Julia module using the following command.

module load julia

By default, it loads the latest stable version.

Using Julia on the command line

After loading the Julia module, we can use Julia with the julia command. Without arguments, it starts an interactive Julia REPL.

julia

For available command line options, we can read the manual.

man julia

The official Julialang documentation or the discourse can answer most questions regarding the features of the Julia language. The Julia language includes the standard language features in Base. Additionally, it includes various packages in the Julia installation as part of the standard library. Julia's REPL and Pkg, the package manager, are two important packages within the standard library. Pressing ] in the Julia REPL will allow access to the package manager's REPL. The Pkg documentation provides more information on how to use Julia's package manager.

More information

We can print details about paths and environment variables set by the Julia module as follows.

module show julia

We can also print useful information about the Julia version, platform and environment in the REPL as follows.

using InteractiveUtils  # automatically load in the REPL
versioninfo()

Using environments

Julia manages dependencies of projects using environments. An environment consists of two files, Project.toml and Manifest.toml, which specify dependencies for the environment. We define project metadata, dependencies, and compatibility constraints in the Project.toml file. Adding or removing packages using the package manager manipulates the Project.toml file in the active environment. Furthermore, the package manager maintains a full list of dependencies in the Manifest.toml file. It creates both of these files if they don't exist. Let's consider a Julia project structured as follows.

project/
├── script.jl
├── Project.toml
└── Manifest.toml

We can activate an environment using the --project option when starting Julia or use the Pkg.activate function in the existing Julia session. For example, we can open the Julia REPL with the project's environment active as follows:

julia --project=.

We can call the Base.active_project() function to retrieve a path to the active project, that is, Project.toml file.

Activating an environment does not automatically install the packages defined by Manifest.toml or Project.toml. For that, we need to instantiate the project as follows:

using Pkg
Pkg.activate(".")
Pkg.instantiate()

Alternatively, we can use the following one-liner:

julia --project=. -e 'using Pkg; Pkg.instantiate()'

Now, we can run the script using the project's environment as follows:

julia --project=. script.jl

Julia will activate the default environment if we don't specify an environment. Preferably, we should use a unique environment for Julia projects instead of the default environment. That way, we can manage the dependencies of different Julia projects separately.

Adding packages to an environment

On the Julia REPL, we can use the package manager by importing it.

using Pkg

We can activate a Julia environment on the current working directory as follows.

Pkg.activate(".")

We can add packages to the active environment using the Pkg.add function. For example, we can add the ArgParse package as follows.

Pkg.add("ArgParse")

Furthermore, we can set compatibility constraints to a package version. For example, we can add compatibility to ArgParse as follows.

Pkg.compat("ArgParse", "1.1")

We can also set compatibility constraints to the Julia version.

Pkg.compat("julia", "1.8")

Code loading and the shared environment

The Julia constants Base.DEPOT_PATH and Base.LOAD_PATH constants control the directories where Julia loads code. To set them via the shell, we use the JULIA_DEPOT_PATH and JULIA_LOAD_PATH environment variables. We can call the Base.load_path() function to retrieve the expanded load path. The Julia module automatically appends the default depot and load paths to ensure the standard library and shared depots are available.

The first directory on the depot path controls where Julia stores installed packages, compiled files, log files, and other depots. We can change the directory by prepending the JULIA_DEPOT_PATH with a different directory. For example, we can use the following by replacing the <project> with your CSC project.

export JULIA_DEPOT_PATH="/projappl/<project>/$USER/.julia:$JULIA_DEPOT_PATH"

Changing the default depot directory.

By default, the first depot directory in the depot path is $HOME/.julia. However, the home directory has a fixed quota for Puhti and Mahti. Therefore, we recommend changing the directory to a directory under Projappl or Scratch to avoid running out of quota because some packages install a large number of files. Afterward, you can safely remove the default depot directory using rm -r $HOME/.julia.

The CSC-specific shared depots are installed in the CSC_JULIA_DEPOT_DIR directory, and the shared environment is in the CSC_JULIA_ENVIRONMENT_DIR directory. We can look up the shared packages and their versions using the package manager as follows:

julia --project="$CSC_JULIA_ENVIRONMENT_DIR" -e 'using Pkg; Pkg.status()'

Creating a package with a command line interface

We should package the code as a code base grows instead of running standalone scripts. A Julia package includes a module file, such as src/Hello.jl, and the Project.toml file. Including a command line interface in your program, such as src/cli.jl, is also wise. Let's consider a project structured as below.

Hello.jl/         # the package directory
├── src/          # directory for source files
│   ├── Hello.jl  # package module
│   └── cli.jl    # command line interface
└── Project.toml  # configurations and dependencies

The Project.toml file defines configuration and dependencies like the following example.

name = "Hello"
uuid = "d39f8c29-790d-4dca-9a6b-e0bca2099731"
authors = ["author <email>"]
version = "0.1.0"

[deps]
ArgParse = "c7e460c6-2fb9-53a9-8c5b-16f535851c63"

[compat]
julia = "1.8"
ArgParse = "1.1"

The src/Hello.jl file must define the module keyword with the package name. It also exports the functions and variables we want to expose in its API. For example, the Hello module below defines and exports the say function.

module Hello

say(s) = println(s)

export say

end

We can use the ArgParse package to create a command line interface src/cli.jl for the package. For example, the command line interface below defines an option --say whose value is parsed into a string and supplied to the say function imported from the Hello module.

using ArgParse
using Hello

s = ArgParseSettings()
@add_arg_table! s begin
    "--say"
        help = "say something"
end
args = parse_args(s)

say(args["say"])

We can use the command line interface as follows.

julia --project=. src/cli.jl --say "Hello world"

We should define and use a command line interface because it is more flexible than hard-coding values to the scripts.

Running Julia jobs on Puhti, Mahti, and LUMI clusters

We explain how to run serial, parallel, and GPU jobs with Julia on Puhti, Mahti, and LUMI in the Running Julia jobs on Puhti, Mahti, and LUMI clusters tutorial.

Using Julia on Jupyter and VSCode

Julia is also available on the web interface via Jupyter and VSCode.