Advanced usage

For advanced users it is suggested to install git, bash and vscode or vscodium in addition to Julia. vscode and vscodium both have a very good plugin for Julia support, see https://www.julia-vscode.org. Installation instructions: Julia and VSCode .

Whe using vscode, I do NOT use the Julia terminal provided by vscode, but the normal bash terminal which is also available in vscode by selecting Terminal->New Terminal From this terminal I start Julia with julia --project or a different command as explained below. This makes it easier to understand what happens and is also faster when you need to restart.

Forking the repository and creating a custom system image

To reduce the startup time it is suggested to use a custom system image that contains all the packages you use on a daily base in compiled form.

Go to the website https://github.com/ufechner7/KiteModels.jl and click on the Fork button at the top right.
clone the new repository which is owned by you with a command similar to this one: git clone https://github.com/ufechner7/KiteModels.jl Your own git user name must appear in the URL, otherwise you will not be able to push your changes.

After cloning the repo you can create a new system image:

cd KiteModels.jl
cd bin
./create_sys_image --update

This will take about 6 min on a Ryzen 7950X CPU. You should now see a new file in the bin folder:

~/repos/test/bin$ ls -lah kps*
-rwxrwxr-x 1 ufechner ufechner 723M apr 18 18:23 kps-image-1.10-main.so

You can launch julia such that it makes use of this system image with the commands:

cd ..
./bin/run_julia

If you now run any of the examples the time-to-first-plot (TTFP) should be less than 10s:

julia> @time include("examples/simulate_simple.jl")
lift, drag  [N]: 597.47, 129.29
Average number of callbacks per time step: 83.8866
  9.370429 seconds (17.95 M allocations: 1.359 GiB, 4.26% gc time, 50.76% compilation time: 26% of which was recompilation)

julia>

A second run of this command needs about 3.6 s which means the startup time (load and compilation time of the package and the libraries) has been reduced to about 5.77s.

Without a system image the first time execution of the script "simulate_simple.jl" on the same computer is about 16.4 seconds while the time for the second execution is the same (3.6s). So now about 7s of time are saved after each restart.

Hints for Developers

Coding style

add the packages TestEnv and Revise to your global environment, not to any project
avoid hard-coded numeric values like 9.81 in the code, instead define a global constant G_EARTH or read this value from a configuration file
stick to a line length limit of 120 characters
try to avoid dot operators unless you have to.

Bad: norm1 .~ norm(segment) Good: norm1 ~ norm(segment)

if you need to refer to the settings you can use se() which will load the settings of the active project. To define the active project use a line like set = se("system_3l.yaml") at the beginning of your program.
use the \cdot operator for the dot product for improved readability
use a space after a comma, e.g. force_eqs[j, i]
enclose operators like + and * in single spaces, like 0.5 * (s.pos[s.num_C] + s.pos[s.num_D]); exception: mass_tether_particle[i-1]
try to align the equation signs for improved readability like this:

    tether_rhs        = [force_eqs[j, i].rhs for j in 1:3]
    kite_rhs          = [force_eqs[j, i+3].rhs for j in 1:3]
    f_xy              = dot(tether_rhs, e_z) * e_z

Outlook

The next steps:

re-implement the KPS4 model using ModelingToolkit
add a Matlab/ Simulink wrapper similar to the Python wrapper pykitemodels