This is quite a long post on the topic of working directory and (Neo)vim.

There will be a second part to this story, and maybe a TLDR; if you ask nicely :)

Introduction: what is the cwd ?

In a command line prompt

cwd is short for “current working directory”.

Every command you run has its own current working directory. When you start a terminal emulator, your first cwd is your home directory (/home/user), and then you can use cd to change the working directory.

At any time you can display your working dir by typing pwd, and usually your prompt is configured to give you this information.

Here the prompt is configured to display the working directory between square brackets:

[/home/user] $ pwd
[/home/user] $ cd /foo/bar
[/foo/bar] $ pwd

Setting the working directory allows you (among other things), to type relative paths instead of full paths.

For instance, let’s assume you have some C code in /path/to/foo/src, you need to edit the source code for bar and its header.

You could run:

[/home/user] $ vim /path/to/foo/src/bar.c
[/home/user] $ vim /path/to/foo/src/bar.h

But it’s much more convenient to use:

[/home/user] $ cd /path/to/foo/src
[/path/to/foo] $ vim bar.c
[/path/to/foo] $ vim bar.h

In Vim

Vim is no different. When you start vim, it gets the working directory of your shell. And then you can type commands like :e to open paths relative to your working directory.

Using the same example, after:

[/home/user] cd /path/to/foo/src
[/path/to/foo] vim

you can run :e bar.c, and then :sp bar.h

The problem: working with several directories

This is all well and good, but what happens when you start working on several projects ?

For instance, you could be working on the HTML documentation of your project, in /path/to/foo/doc.

You need to see the .html and .css files when you are editing the documentation, but also sometimes you want to have a look at the actual code.

An obvious solution is to create a new tab, with :tabnew doc, but then if you want to edit index.html you have to type :e ../doc/index.html.

An then if you want to edit the CSS you have to run: :vs ../doc/style.css

So you have to keep typing ../doc/ and it’s annoying.

My journey to the prefect workflow

I’ve got this issue for years. It’s taken me a long time to find a solution for this problem, so I thought I’d share this process with you.

Step 1: using autochdir

Vim has an option for this. Here’s the documentation:

'autochdir' 'acd'   boolean (default off)
    When on, Vim will change the current working directory whenever you
    open a file, switch buffers, delete a buffer or open/close a window.
    It will change to the directory containing the file which was opened
    or selected.
    Note: When this option is on some plugins may not work.

That was my first try.

I think it’s not a good solution (and not only because it’s what Emacs does this by default :P)

Here’s why.

Let’s assume your project became more complex, and you start having a subproject called baz.

Here’s what your source code looks like:


When you are editing bar.h, you can type :e baz/baz.c and it feels natural.

But then, if you want to go back from baz/baz.c to bar.h, you have to use :e ../bar.h which feels strange…

Worse, let’s assume you have:

/* in baz/baz.h */
#include <bar.h>

You may want to open bar.h by using gf, or auto-complete the path to the header using CTRL-X CTRL-F, but you can’t since you don’t have the correct working directory!

Plus the doc says it may break some plugins…

Step 2: using :cd

Vim has a command to change the working directory as well.

So back to our example, you can do:

:cd /path/to/foo
:cd src
:e bar.c
:e baz/baz.h
:cd ../doc
:e index.html

Well that’s much better! There’s still a problem though: :cd changes the working directory for the whole vim process.

So if you run tabprevious to go back editing the C code, your working directory is no longer correct, and you have to re-type :cd src.

Step 3: using :lcd

Luckily, vim has a command to change the working directory just for the current window: :lcd. So I started using that.

And then I realized I often started vim directly from my home directory, so I had to type things like:

:e /path/to/foo/src.c
# Ah, I need to change the working directory...
:cd /path/to/foo/

That’s awful. You type the same path twice!

Or I used to type:

:cd /path/to/foo/src
:e foo.h
# Time to fix the doc
:tabnew ../doc
:cd ../doc
# Ah crap, I meant :lcd ...

Step 4: using a custom command

I don’t recall how I found it, but here’s what has been in my .vimrc since quite some time:

" 'cd' towards the dir in which the current file is edited
" but only change the path for the current window
map <leader>cd :lcd %:h<CR>


  • map defines a new command
  • leader is replaced by what you set with let mapleader. Default is backlash, but you can use any character for this.
  • lcd is the command we just talked about
  • % represents the current file, and what’s after the : is called a “filename modifier”
  • h is a filename modifier corresponding to the “dirname” of the file

You can see the full list of filename modifiers with :help filename-modifiers, and to use them from vimscript you can use the fnamemodify() or expand() functions.

Well, that’s much better. You can start opening a long path, and then change the working dir without retyping all the path components.

Also, you are always using :lcd, so you never change the path globally.

This quickly became the shortcut I could no longer live without…

Step 5: using <leader>ew

This is another trick you can use when you know are going to edit a file that is “near” the file you are currently editing, but don’t want to change the working directory at all.

The code looks like this:

" Open files located in the same dir in with the current file is edited
map <leader>ew :e <C-R>=expand("%:p:h") . "/" <CR>


  • <C-R>= is short for Ctrl-R followed by the equals sign. It allows to enter a vim expression.
  • expand(%:p:d): we see our % friend, which still represents the current filename
  • :p:h: two file modifiers: one to get the full path (:p), and the other to find the dirname (:h)
  • Then we add a / so that we can start typing the filename right away.

Here’s how you use it

:e /some/long/path/to/foo.c
<leader>ew foo.h
# opens /some/long/path/to/foo.h

Step 6: using :TabNew

After a while, I realized I really liked having one working directory per tab. I found myself typing stuff like:

:tabnew /some/path
:e some-path-at-the-top
:e subdir/otherfile.c

I was opening a file at the top of the project just to be able to use my <leader>cd command and start thinking of better ways.

So finally I came up with a new command:

" Change local working dir upon tab creation
function! TabNewWithCwD(newpath)
  :execute "tabnew " . a:newpath
  if isdirectory(a:newpath)
    :execute "lcd " . a:newpath
    let dirname = fnamemodify(a:newpath, ":h")
    :execute "lcd " . dirname

command! -nargs=1 -complete=file TabNew :call TabNewWithCwD("<args>")

Hopefully by now you should understand what this does: I create a function that calls fnamemodify to get the dirname of the file I want to open in a new tab, and then calls :lcd with the correct argument.

Step 7: Switching to Neovim

And then I switched to Neovim, and two things happened:

  • 1/ Neovim folks added the TabNewEntered event
  • 2/ They also added the tcd command, which allows to change the working dir just for a tab.

That’s exactly what I needed!

So now my vimrc looks like:

function! OnTabEnter(path)
  if isdirectory(a:path)
    let dirname = a:path
    let dirname = fnamemodify(a:path, ":h")
  execute "tcd ". dirname

autocmd TabNewEntered * call OnTabEnter(expand("<amatch>"))

So no matter how I enter a new tab, my working directory is automatically set to the correct location, and when I switch tabs I switch working directories too. Perfect!

Step 8: Fixing vim-fugitive

To conclude my journey, I’d like to share just a tiny bug fix I had to do for using vim-fugitive

It occurred when I was working in a git project, with the Python code in python/, and the documentation in doc/.

I was in my python/ tab, ran :Ggrep and suddenly I got results for all the files in the repositories, including the documentation.

That’s because :Ggrep starts by changing the working directory to the top directory of the current git repository.

Fortunately the fix was easy, here’s my pull request on github.

I don’t know if it will be accepted by Tim Pope, but I hope that now you understand why this is such a big deal for me :)

Thanks for reading!