Last month marked four years since I started my dotfiles repository on GitHub. Now it's in a state of disrepair! I had no idea that I hadn't made a single commit to it throughout 2017. I did though, I just didn't push them. There are local changes on every computer that I work with. Some were the results of experiments I later forgot about and others were quickly-needed hacks that I didn't find time to refactor and stabilise. Now it's time to clean it up.
I came across a video which explains some lesser-known Vim tips. With just a few lines of Vimscript, you can moderately match the functionality of some common Vim plugins.
The first tip is to add
** to the search path when using Vim's
command. This gives behaviour similar to a fuzzy finder but with the built-in,
kind-of awkward-for-long-lists wildmenu. I think this has to potential to be
used instead of a fuzzy finder for small code bases that have short-ish file
names. If you're working with a monolith, maybe not. Fuzzy finding with the
first letter of each camel-cased word in a class name can be really helpful for
The second tip is explaining Vim's built in ctags integration for jump-to-definition functionality. Code completion is covered separately in the video but ctags supports this as well. Unfortunately it seems like most effort in this space is now being dedicated towards language server protocols. The Rust Language Server is an implementation of the protocol and has support from the core team.
Build integration is explained with Vim's
makeprg configuration variable. This
is the external program called when invoking
make from inside Vim. For simple
cases it's perfectly sufficient. The video goes on further to explain how to
integrate marked errors with Vim's quickfix list so that you can easily navigate
them. In general, this needs to be done on a per-language basis, so I agree with
the speaker that plugins are useful here.
Personally, I've never seen the need for snippets. Maybe I'm just using the wrong languages.
That leaves file browsing. The speaker talks about netrw, Vim's built-in file browser. Unfortunately, netrw isn't the most stable piece of software. This Reddit thread talks about some of the alternatives. NERDTree is the most frequently recommended and the most popular “project draw”-type plugin. Such plugins are an ongoing point of contention within the community, with those that want to make Vim IDE-like and those that say this is unidiomatic and that Vim should stick to traditional Unix philosophy: do one thing and do it well. For Vim, that's editing text. As such, simpler alternatives have been created like FileBeagle and vim-dirvish.
Deciding what Vim should be
If there's three camps of coders, you can bet there's be multiple views on idiomatic Vim configurations. Arguing about it online is all well and good to see different opinions and help you decide which camp you sit in. Vim, along with its plugin ecosystem, is completely open and that means you can configure out however you please.
For me, I want Vim to be fast. Thinking back on my usage in the last four years, I'm comfortable falling back into the terminal. I used very few of the plugins I had installed to their full potential. And yet, seeing that delay when opening Vim over and over again would agitate me.
Realising that removes the need for many plugins. Most of us probably use Git everyday. Tim Pope's fugitive.vim tops many “must-have Vim plugins” lists. The problem is that the time you spend interacting with Git is not that much compared to the time spent reading and writing code. I interact with Git a few times per day on average, maybe ten times, but I'm editing code far more frequently than that. Enabling it inside Vim seems unnecessary. Sure it might be helpful to see which branch you're on, but do you change branches that frequently? It could be useful to see the modifications in the sidebar like vim-gitgutter offers, but how often do you actually make decisions based on that?
Instead, I think it makes more sense to use a dedicated tool like tig. You can think of this like the mode-based philosophy of Vim, but more meta. We were in editing mode and now we're in review mode, where we check our changes before committing them. This is the workflow I was already using without realising it, so it's a natural fit.
We can take this further. During the orientation of learning a new codebase, exploring with a file browser is useful. Once you're familiar with the codebase though, you'll often just want to jump to specific files. You have a mental map of the codebase in your head. Why include a file browser then? Instead, we can again use an external tool like ranger!
It's likely there are other phases in our workflows that we can delegate to a specialised tool. Both tig and ranger have Vim integrations available and delegating to them when they are needed feels natural. It keeps Vim snappy and focussed on what it's good at.
Less plugins, more configuration
Plugins are designed to be applicable to as many people as possible. They will support use cases you may never need. Hence I am also trying to instead take ideas from the plugins I like and configure them inside Vim myself. It's a great way to learn Vimscript, offers me tailored control on the behaviour, and makes sure that every time I benchmark Vim's startup time I know why it's behaving the way it is.
Modern Vim is due for release next month. I loved Drew Neil's previous book, Practical Vim. Having read it when I'd been using Vim for only a few months, much of it went over my head, but it convinced me that the methodology works. I will tinker with my configuration for now in the hopes that Modern Vim will soon enlighten me much like Practical Vim did before it.