Developing Persefone

Setting up

If you haven't worked with Julia before, here are detailed instructions for how to set up your development environment. The main development is currently done on Linux (and as the primary execution platform will be an HPC, Linux compatibility is important), but developing on Windows works too.

Visual Studio Code on Windows

  1. Download and install Julia, git and Visual Studio Code.

  2. Install the Julia extension for VS Code: In VS Code, open the extensions pane (Ctrl+Shift+X). Search for and install Julia Language Support.

  3. Clone the Gitlab repository: In VS Code, open the source control pane (Ctrl+Shift+G). Click on Clone and enter the repo URL. Then select a folder on your computer to download the files into, and let VS Code open the project once it has been cloned.

  4. Start a Julia REPL: In VS Code, bring up the command palette (Ctrl+Shift+P). Execute the command Julia: Start REPL. Then install all dependencies of Persefone by running using Pkg; Pkg.activate("."); Pkg.instantiate(). (This will take some time.)

  5. Open the file run.jl and click Execute (triangular button in the top right). The source code will compile (this can take a lot of time the first time you do it) and run a default simulation.

  6. Further steps: You may want to familiarise yourself with how to use git with VS Code. You may also want to clone the Persefone Desktop repository (repeat steps 3 to 5).

Emacs on Linux

You can of course also use VS Code on Linux. In that case, follow the instructions above.

Make sure you have git and Julia installed. Git should be in your distro's repos (e.g. sudo apt install git). To install Julia, download the binary and unpack it. For greater ease of use, copy the unpacked files to /usr/local/lib/julia (or similar) and create a symlink to the executable: sudo ln -s /usr/local/lib/julia/bin/julia /usr/local/bin/julia. Then go the to folder that you want to use for development and run git clone https://git.idiv.de/persefone/persefone-model.git . in your terminal.

There are a couple of addons that make working with Julia much nicer in Emacs:

  1. julia-mode gives syntax highlighting. Install with M-x package-install julia-mode.

  2. julia-snail provides IDE-like features, especially a fully-functional REPL and the ability to evaluate code straight from inside a buffer. Note that the installation can be somewhat tricky. You first need to manually install all the dependencies of its dependency vterm, then install vterm itself with M-x package-install vterm, before you can do M-x package-install julia-snail. Then add it to your init.el with (require 'julia-snail) and (add-hook 'julia-mode-hook #'julia-snail-mode).

  3. company-mode integrates with Snail to give code completion. Install with M-x package-install company, then add (add-hook 'julia-mode-hook #'company-mode) and (global-set-key (kbd "C-<tab>") 'company-complete) to your init.el.

  4. magit is a great git interface for Emacs. Install with M-x package-install magit and add (global-set-key (kbd "C-x g") 'magit-status) to your init.el.

Development workflow

  1. Pull the current version from the master branch on Gitlab: https://git.idiv.de/persefone/persefone-model.

  2. If you are working on a new feature, create a new branch to avoid breaking the master branch. (The master branch on Github should always be in a runnable and error-free state.)

  3. Implement your changes.

  4. Run an example simulation and the test suite to make sure everything works without crashing (make run and make test on Linux, or execute run.jl and test/runtests.jl manually.)

  5. Commit your work frequently, and try to keep each commit small. Don't forget to add relevant tests to the test suite.

  6. Once your satisfied with your work, do another pull/merge from the master branch in case somebody else changed the branch in the meantime. Then merge your work into master and push to the Gitlab server.

  7. Repeat :-)

The Gitlab issue tracker can be used to create, discuss, and assign tasks, as well as to monitor progress towards milestones/releases. Once we have a first release, we will start using semantic versioning and a changelog.

Important libraries

Revise.jl

Revise.jl allows one to reload code without restarting the Julia interpreter. Get it with Pkg.add("Revise"), then add using Revise to .julia/config/startup.jl to have it automatically available.

Test

Persefone uses the inbuilt Julia testing framework. All new functions should have appropriate tests written for them in the appropriate file in the test directory. (See test/runtests.jl for details.) There are three ways to run the test suite: in the terminal, executing make test or cd test; julia runtests.jl; or in the Julia REPL, Pkg.activate("."); Pkg.test().

Documenter.jl

The HTML documentation is generated using Documenter.jl. Therefore, all new functions should have docstrings attached. New files need to be integrated into the relevant documentation source files in docs/src, and if necessary into docs/builddocs.jl. To build the documentation, run make docs, or cd docs; julia builddocs.jl (if using the latter, don't forget to update the date and commit in docs/src/index.md).

Graphics and user interface

Persefone uses Makie as a plotting library to generate its output graphics. Additionally, Persefone Desktop uses QML.jl to create its graphical user interface.

Unitful.jl

Throughout the source code, variables can be tagged with their appropriate units using the Unitful.jl library. This makes the code easier to understand, and also allows automatic unit conversion:

julia> 1ha == 10000m²
true

julia> 2km |> m
2000 m

julia> 2km / 10m
200.0

Within Persefone, the following units and dimensions have been imported for direct usage: cm, m, km, , ha, km², mg, g, kg, Length, Area, Mass.

Dates

Persefone expands the default Dates library with the AnnualDate type, which can be used to store dates that recur every year (e.g. migration or harvest). AnnualDates can be compared and added/subtracted just as normal dates. Use thisyear() to convert an AnnualDate to a Date.