Codi.vim + fullscreen buffer

Recently I discovered codi.vim on vimawesome.com. It sells itself as an "interactive scratchpad for hackers". After playing around with it for 5 minutes I was completely sold.

I pair program on a weekly basis using various different languages. I used to turn to a REPL for help but this quickly becomes annoying when trying to write multi-line expressions. Making a typo in the middle of a do ... end block in Ruby is quite costly:

Another thing to note here is that code and results are printed sequentially. This means that any code you write goes off-screen at least twice as fast because the result is printed below it.

In case of errors and stack traces it can become a lot worse as well. Codi solves all of this by allowing you to type in one pane and it prints the results in a separate pane side-by-side:

This feature, and the fact that I can toggle this buffer for multiple programming languages using a single mapping blew my mind but as with anything, it requires some more battle testing to see if it works in common scenario's.

Battle testing

This is where I found out that when used with a split, it will not be fullscreen but rather split the current buffer:

At this point I started experimenting with a small vimscript snippet in my .vimrc. Excluding comments and blanks it is 10 LoC. It wraps the default Codi command in a function which creates an empty buffer in a new tab with the right filetype set before finally running Codi itself. This creates the fullscreen split that I wanted:

Problem solved, or maybe not?

All good so far, but upon closing the buffer, the contents of the buffer are deleted. While automatically clearing the scratchpad is nice, sometimes I want to go back and forth between the code and the scratchpad to add additional code or compare the existing code with the scratchpad code.

This required a second iteration of the original vimscript snippet. This time I did not want to delete the buffer but hide it instead, keeping the code that was already written in it. After hiding the buffer, its tabpagenr() is stored in a dictionary keyed by the filetype of the buffer.

Then, when toggling codi for a filetype that is stored, switch to that tab instead of opening a new one (there is even a check if that tab actually exists):

And there we go! Aside from the fact that the buffer now remains, the name is also scratch::ruby. This is a bit more descriptive than having just [Scratch] in :ls output when multiple codi buffers of a different filetype are active.

Let's wrap it in an issue

It was a fun experience tweaking this plugin to my liking. I thought this feature would be so awesome that I even proposed it in an issue. Unfortunately it didn't make it since it was too workflow specific.

As you may have been able to see in the issue though, the code in that snippet is slighty different from the example gist linked in this post. The main difference is this line which resizes codi to 50% of the buffer width whenever a new instance is spawned:

" since it is fullscreen, I'd like a 50/50 splitlet g:codi#width = winwidth(winnr()) / 2

The author thought it would be a nice addition to have percentage width support and for the snippets to live elsewhere for others to use in their .vimrc. So I went ahead and wrote a PR for that instead, this is why you see let g:codi#width = 50.0 in the gists :)

Conclusion

All in all it was a great experience discovering this plugin. It is like one of those rare cases with a song where after a few seconds you just know "this song is awesome". I believe I'll be using it for a long time to come for quickly testing and showing some code.

This also gave me the chance to write my first non-organization bound PR for a project which I like and now also use. Something which I wanted to do for a long time but couldn't really find the right project to contribute to. I want to thank Ethan Chan for his awesome plugin and for being a kind and friendly maintainer!

That's it! If you want a fullscreen workflow with codi.vim, you have it. If you want to persist the code in the buffer, you have it.

Until next time.

👋