Updating and managing multiple Julia versions

Out of the box, Julia has a great package manager but it’s missing something like conda or pyenv to manage different versions of Julia, or to update your current installation of Julia to the latest release. See below for what I use on my Mac:

Feb 2023 Update:

I recently switched out of conda and into pyenv and poetry on my mac, which ended up breaking my previous method of using pyjill to manage my julia installation (which requires python).

This was a great opportunity to switch to juliaup, which is written in Rust and offers much of the same functionality without the dependence on python. However, it’s not perfect (particularly to users in PRC, as the author of pyjill notes here) and I’ll be keeping the old instructions of using pyjill for now:

Click here for the old pyjill instructions (Python required)

Step 1: Use `jill`/`pyjill` to Download & Install Julia

I used to mess with downloading the binaries and install files directly from the julialang site and also tried homebrew, but both of these options were really clunky when it came to updating your existing installation. Instead, I just use pyjill (Github), which works perfectly on both OSX, Windows, and Linux (note: requires python >=3.6).

# Install or update the latest version of jill (requires python>=3.6)
pip install jill --user -U
# export the jill PATH somewhere, if this is your first time installing jill
echo 'export PATH=$PATH:/home/$USERNAME/.local/bin' >> ~/.zshrc
source ~/.zshrc
# Install the latest stable release
jill install

Then the next time you need to update Julia, you only need to do jill install to get the latest.

Step 0: First-time Install with `juliaup`

The instructions on the juliaup repo are extensive, but I’ll share a few key commands here:

On a Mac, it’s advised (as of Feb’23) to use the following curl command instead of brew

curl -fsSL https://install.julialang.org | sh

This will run an interactive install script and for convenience it will update your PATH and install the latest release and set it as your default if you don’t have Julia already installed.

Then juliaup status will show you the list of versions you have installed, and show you which is the default version of julia that’s symlinked to julia.

juliaup status

Switching environments

You can always install an older release, let’s say you want 1.8.4:

# Add an older version of Julia, let's say 1.8.4
juliaup add 1.8.4
# Then to run that version of julia, all you would need to do is add `+1.8.4` when you launch julia, i.e.:
julia +1.8.4

You can also set it as the default channel with juliaup default:

juliaup default 1.8.4

Step 1: Updating to the latest release with `juliaup`

juliaup will automatically check for the latest release every so often when you launch julia, but you can manually update with:

juliaup update

Step 2: Remove the Old `.julia` Folder (optional, linux/OSX instructions)

Another time sink I had was realizing that with a fresh update, I couldn’t add packages anymore to julia because I didn’t have root access to a folder hidden deep within the .julia folder in my user directory. I had to delete it and then julia worked fine after that after rebuilding the registry. This might not be as much of an issue after Julia 1.7, where they changed the manifest and how the package registry handles different Julia versions.

# Go to your home directory
cd
# Delete that sucker with prejudice
sudo rm -rf .julia

Step 3: Update Jupyter Kernels

Next we’ll need to remove the old julia kernel from Jupyter with jupyter kernelspec remove {{ kernel name }}:

# See the list of existing jupyter kernels and find your old install
jupyter kernelspec list
# My previous one was called julia-1.6, let's remove it
jupyter kernelspec remove julia-1.6

Then use IJulia to add a new one. Whenever you change the Julia binary you need to have IJulia rebuild in order to register the new kernel to jupyter, so go back to Julia and go to the Pkg prompt:

# In the Julia REPL, press the ] key to bring up the Pkg prompt
(@v1.8) pkg> add IJulia
(@v1.8) pkg> build IJulia

At this point, if you’re upgrading from Julia 1.7+ and it’s a minor update (i.e. 1.7.2 to 1.7.3), your base package manifest should still be there and Julia will precompile and build those other packages so they’ll be ready to go - no need to re-add them.

Step 4. Update Your Libraries (if upgrading from Julia < 1.7)

From here on out you’ll need to probably build and update your other libraries too, and if you use stuff like PyCall(link) where it downloads a distribution of python and hides it in your julia folder, have fun managing all of that! I’ll do a writeup later when I figure this out a little more.

# A few common libraries that I install for data exploration and analysis
(@v1.8) pkg> add Revise Plots Gadfly DataFrames DataFramesMeta XLSX CSV RDatasets Parquet

Updating and managing multiple Julia versions

Step 1: Use jill/pyjill to Download & Install Julia

Step 0: First-time Install with juliaup