I really want to like Julialang. They’re making some serious promises with what they can do, and #IWantToBelieve. However, their tutorials though… rough. So this post (or series of posts) is a chronicle of my desperate simping to make Julia work, and if it can work for me (Data Science, Visualization, and Graph Analytics).
Why Julia?
I use Python3 and Jupyter Notebook a lot, but the Notebook dies when I start bringing visualization and NetworkX (graph analytics) to the mix. So I’m hoping that the speed, visualization, and multiprocessing promised by Julia help me process datasets without resorting to setting up a big data analysis solution.
So why are the tutorials rough?
Mostly inconsistency due to constant updating. Some is avoidable some not so much. For example:
- The tutorials recommend installing Julia 1.0 LTS, but then do the tutorials with Julia 1.4.
- Julia Academy has you install JuliaPro and nteract, but the tutorials have you use Jupyter Notebook.
- Some packages require installed programs and environmental variables setup. Like RCall requires R to be installed and $R_HOME to be set.
Installation/Setup
I tried a bunch of permutations of the install instructions, and this is what I use:
Note: I am using a Macbook with the Apple Silicon M1 chip.
- I already had Anaconda and XCode installed, so I already had Jupyter Notebook setup.
- I installed Julia v1.5, because 1.0 kept giving me errors.
- I created the symbolic link from the julia bin to /usr/local/bin/ so Julia was in my path
ln -s /Applications/Julia-1.5.app/Contents/Resources/julia/bin/julia /usr/local/bin/julia - launched Julia from my terminal
- press the “]” key which puts you in package mode.
- then entered
add IJuliato add the Julia kernel to Jupyter Notebook. - I also used this terminal to add a bunch of other packages.
add CSV XLSX ScikitLearn DataFrames Plotlyand whatever else you need.- If you got an install error and you fixed the issue, you must build the package instead of adding it again or it will not work. e.g.
build Plotly - If you got an install error and you cannot fix the issue, just remove the package so your kernel wont crash. e.g.
remove TensorFlow
Alternative Setups
If you don’t want Anaconda but still have python3, you can use pip to install Jupyter.
If you don’t want to use Jupyter Notebook, but still want to use the tutorial Jupyter Notebooks you can use nteract.
If you want to use Atom, then I recommend installing JuliaLang, then install Atom, then install the Juno addons for Atom.
Results
Julia has made some serious promises that have definitely peaked my interests. However, the inconsistent install methods, tutorial versions, and hidden package dependencies makes Julia difficult for newbies. I will let you know if it was worth all the trouble in another post.
Links:
Julia Lang site: https://julialang.org/
Julia Tutorials:https://juliaacademy.com/
Julia Youtube: https://www.youtube.com/user/JuliaLanguage
JuliaPro IDE: https://juliacomputing.com/products/juliapro/
Anaconda: https://www.anaconda.com/products/individual
nteract: https://nteract.io/kernels/julia
