Skip to content

Latest commit

 

History

History
43 lines (33 loc) · 1.69 KB

README.md

File metadata and controls

43 lines (33 loc) · 1.69 KB

Docker for Julia

In this directory, the followings are prepared.

  • Package management with the default Julia function.
  • Multi-thread setting.
  • Use of PrecompilePackage.

Package management and PackageCompiler setting.

Julia programming language provides the package management system. It is simple to maintain the package dependencies by files Manifest.toml and Project.toml. Therefore, in Dockerfile, we do not explicitly write the package installation commands.

First, comment out all codes below ENV JULIA_PROJECT=/workdir since we do not have Manifest.toml, Project.toml or other files when you prepare the contents.

Then, launch up the container and run the code

using Pkg
Pkg.activate(".")
Pkg.add("FreqTables")

Pkg.add("PackageCompiler")
using PackageCompiler
create_sysimage(["FreqTables"], sysimage_path="freqtables.so")

which generates Manifest.toml and Project.toml. For the second run, you can comment in code which previously is commented out. Then other people can reproduce your results by docker compose up.

The last bunch of code are for creating sysimage. Then, you comment in codes, and can use Julia with precompiled packages.

Note: You have to COPY files such as "Manifest.toml", "Project.toml", and "freqtables.so" to workdir directory in Dockerfile, since volume mounting is conducted after you launch the container. It means when you create a docker image, the workdir directory is empty!

PackageCompiler.

Taking long time to load the package is serious issue in Julia programming. At least, we want to precompile graphical pacakges such as CairoMakie, Plots, and Gadfly.
For illustration purpose, we precompile FreqTables and use in Jupyter lab.