For some of our projects at work we have decided that we want to stick with Node’s LTS releases for the next couple of months. In order to not prevent experimentation with newer versions must of us have NVM installed.
In order to make switching between versions easier depending on what project I’m
currently working on, I’m playing a little bit around with a little shell helper
that should load the correct version for a project automatically from a
As this is also the first time I’m trying to extend my shell since switching to fish I’ve learnt quite a bit. One thing in particular I’ve found quite interesting: Functions support acting as “event handlers” out of the box. I use the term “event” loosely here. To be more concrete, I’m talking about things like signals, actual events, and changes to the value of a variable.
The latter is what I was looking for for my little task at hand. If you declare
a function with the
--on-variable NAME parameter, any change to the variable
will trigger that function.
function hello_dir --on-variable PWD echo "Welcome to $PWD" end
The function above (if you put it for instance in your
~/.config/fish/fish/config.fish file) will echo a nice welcome message
whenever you enter a new directory. That’s because, the global
environment variable has been set to a new value.
In other shells like Bash or ZSH you’d override the
cd function for something
like that. Being able to simply add additional handlers on top of each other is
so much nicer!
But, as mentioned above, fish’s function command can do even more. Want to do
something when a process signals for instance HUP?
--on-signal SIGHUP has you
covered. Do you want to get notified when a certain child process exists? Use
--on-process-exit 123 for the process with the PID 123.
As for what fish considers events:
$ function event_handler --on-event CUSTOM_EVENT $ echo "Custom event was triggered with these args: $argv" $ end $ emit CUSTOM_EVENT something Custom event was triggered with these args: something
The shell already emits a couple of internal events, like
fish_postexec. These sound like nice entry points for further exploration 😉
To get back to my original use-case, I’m not completely sure yet, how to implement it. Obviously, I won’t execute nvm with the content of some random file by default but probably implement a little white-listing-logic here. We’ll see 😊