Jumper is a command-line program that helps you jumping to the directories and files that you frequently visit, with minimal number of keystrokes. It relies on fzf for UI and is heavily inspired by z.

It differentiates itself from the plethora of similar tools on the following points:

• Efficient ranking mechanism which combines the "frecency" of the match (as z does) and the accuracy of the match (as fzf or fzy do). This allows to find files/folders accurately in very few keystrokes. More details here.
• "Orderless" search: match a sequence of tokens in any order.
• It is not restricted to folders. It allows to quickly navigate files, or anything you want (you can easily create and query a new custom database).
• It can be run in "interactive mode", relying on fzf for the UI.
• Written in C, for speed and portability.

Usage - Installation - Vim-Neovim

Just move around files and folders as usual, jumper will keep track of them. Then,

• Use z <something> to jump to the most frequent/recent directories matching <something>.
• Use zf <something> to open (in $EDITOR) the most frequent/recent file matching <something>.
• Use Ctrl Y to fuzzy-find directories matching a query interactively (fzf required).
• Use Ctrl U to fuzzy-find files matching a query interactively (fzf required).

All these mappings can be updated, see Configuration below.

The paths that match a given query are ranked based on

• the frecency of the path: how often / recently has this path been visited ?
• the accuracy of the query: how well does the query match the path ?

The ranking of a path at time $t$ is based on the following score

where $\beta = 1.0$ by default, but can be updated with the flag -b <value>. More details about the scoring mechanism are given here.

jumper records visits to files anf direfotries in files whose lines are in the format <path>|<number-of-visits>|<timestamp-of-last-visit>. Given such a database's file, the command

jumper find -f <database-file> -n N <query>

returns the top N entries of the <database-file> (this will typically be ~/.jfolders or ~/.jfiles) that match <query>. Adding the -c flag colors the matched substring. The command

jumper update -f <database-file> <path>

adds the <path> to the <database-file>, or updates its record (i.e. updates the visits' count and timestamp) if already present. From these two main functions, shell scripts (run e.g. jumper shell bash) define various functions/mappings (see Usage above) allowing to quickly jump around.

• Folders: Folders' visits are recorded in the file ~/.jfolders using a shell pre-command. This can be updated by setting the __JUMPER_FOLDERS environment variable.
• Files: Opened files are recorded in the file ~/.jfiles by making Vim run jumper update --type=files <current-file> each time a file is opened. This can be adapted to other editors and the database's file can be updated by setting the __JUMPER_FILES environment variable.

By default jumper uses a simpler version of fzf's "extended search-mode". One can search for multiple tokens separated by spaces. The full fzf-syntax is not implemented yet, only the following token are implemented.

The syntax mode can be changed to fuzzy (use only fuzzy-matches, the characters ^, $ and ' are interpreted as standard characters) or exact (exact matches only), with the --syntax flag.

If the flag -o (--orderless, name coming from emacs' orderless package) is provided, the tokens can be matched in any order. A higher score will be given to matches whose order is closer to the one of the query. More precisely, one adds to the matching_score a term proportional to the number of pairs of tokens that are in the right order (see inversions of a permutations).

The --orderless flag is turned on by default for the z command and interactive searches. This can be changed by editing the __JUMPER_FLAGS environment variable.

By default, matches are "case-semi-sensitive". This means that a lower case character a can match both a and A, but an upper case character A can only match A. Matches can be set to be case-sensitive or case-insensitive using the flags -S and -I.

• A C compiler for installation. The makefile uses gcc.
• Bash (>=4.0), Zsh or Fish.
• fzf. This is not mandatory, but needed for running queries interactively.

You can use the install script to clone and compile jumper set up the shell keybindings automatically:

sh -c "$(curl -s https://raw.githubusercontent.com/homerours/jumper/master/install.sh)"

Alternatively, you can run

git clone https://github.com/homerours/jumper
cd jumper
make install

to compile and move the jumper binary to /usr/local/bin. Then add the following to your .bashrc, .zshrc or .config/fish/config.fish to get access to jumper's functions:

• bash

  eval "$(jumper shell bash)"

• zsh

  source <(jumper shell zsh)

• fish

  jumper shell fish | source

In order to keep track of the visited files, the function jumper update --type=files <file> has to be called each time a file <file> is opened. This can be done automatically in Vim/Neovim, see next section. For other programs, you may want to use aliases (better solutions exist, using for instance "hooks" in emacs)

function myeditor() {
   jumper update --type=files "$1" 
   myeditor $1
}

One typically only needs to add the lines above in ones' .bashrc, .zshrc or .config/fish/config.fish. However, the default keybindings, previewers and "database-files" can still be configured if desired. Here is a sample configuration (for bash)

# Change default folders/files database-files (defaults are $HOME/.jfolders and $HOME/.jfiles):
export __JUMPER_FOLDERS='/path/to/custom/database_for_folders'
export __JUMPER_FILES='/path/to/custom/database_for_files'

# Update jumper's options
# Default: '-cHo -n 500' (colors the matches, replace $HOME with ~/, orderless, see all options by running 'jumper --help')
__JUMPER_FLAGS='-ceo -n 100 --syntax=fuzzy --case-insensitive --beta=0.5'
# Flags for z/zf functions (default: -o):
__JUMPER_ZFLAGS='-eo'

# FZF options for interactive search
# Default: --height=70% --layout=reverse --keep-right --preview-window=hidden --ansi
__JUMPER_FZF_OPTS='--height=30 --keep-right --preview-window=hidden --ansi'

# Change the default binding (ctrl-p) to toggle preview:
__JUMPER_TOGGLE_PREVIEW='ctrl-o'

# Change default files' previewer (default: bat or cat):
__JUMPER_FZF_FILES_PREVIEW='head -n 30'

# Change default folders' previewer (default: 'ls -1UpC --color=always'):
__JUMPER_FZF_FOLDERS_PREVIEW='ls -lah --color=always'

# IMPORTANT: this has to be after the configuration above:
eval "$(jumper shell bash)"

# Change default (ctrl-y and ctrl-u) bindings:
bind -x '"\C-d": jumper-find-dir'
bind -x '"\C-f": jumper-find-file'

Use the function _jumper_clean to remove from the databases the files and directories that do not exist anymore. To clean the files' or folders' databases only, use jumper clean --type=files or jumper clean --type=directories.

This cleaning can be done automatically by setting the variable __JUMPER_CLEAN_FREQ to some integer value N. In such case, the function _jumper_clean will be called on average every N command run in the terminal.

For more advanced/custom maintenance, the files ~/.jfolders and ~/.jfiles can be edited directly.

Querying and updating jumper's database is very fast and shouldn't cause any latency. On an old 2012 laptop, these operations (over a database with 1000 entries) run in about 4ms:

$ time for i in {1..100}; do jumper find --type=files hello > /dev/null; done
real    0m0.432s
user    0m0.165s
sys     0m0.198s
$ time for i in {1..100}; do jumper update --type=files test; done
real    0m0.383s
user    0m0.118s
sys     0m0.209s

For comparison:

$ time for i in {1..100}; do wc -l ~/.jfolders > /dev/null; done
real    0m0.357s
user    0m0.117s
sys     0m0.233s

Jumper can be used in Vim and Neovim. Depending on your configuration, you can either use it

• without any plugin, see below. However, you won't be able to do run queries interactively.
• with the jumper.nvim plugin (prefered). This uses either Telescope or fzf-lua as backend for the UI.
• with the jumper.vim plugin (works for both Vim/Neovim). This uses fzf "native" plugin as UI.

We describe below how to use it without plugins. This only allows to use Z and Zf commands. First, you have to keep track of the files you open by adding to your .vimrc/init.lua

autocmd BufReadPre,BufNewFile *   silent execute '!jumper update --type=files ' .. expand('%:p')

or, if you are using Neovim's Lua api,

vim.api.nvim_create_autocmd({ "BufNewFile", "BufReadPre" }, {
    pattern = { "*" },
    callback = function(ev)
        local filename = vim.api.nvim_buf_get_name(ev.buf)
        -- do not log .git files, and buffers opened by plugins (which often contain some ':')
        if not (string.find(filename, "/.git") or string.find(filename, ":")) then
            vim.fn.system({ "jumper", "update", "--type=files", filename })
        end
    end
})

Then in order to quickly jumper to folders and files, add

command! -nargs=  Z :cd `jumper find --type=directories -n 1 '<args>'`
command! -nargs=  Zf :edit `jumper find --type=files -n 1 '<args>'`

to your .vimrc to then change directory with :Z <query> or open files with :Zf <query>.

You can for instance define a function

fu() {
    RG_PREFIX="jumper find --type=files '' | xargs rg -i --column --line-number --color=always "
    fzf --ansi --disabled --query '' \
    --bind "start:reload:$RG_PREFIX {q}" \
    --bind "change:reload:sleep 0.1; $RG_PREFIX {q} || true" \
    --delimiter : \
    --preview 'bat --color=always {1} --highlight-line {2}' \
    --preview-window 'up,60%,border-bottom, {2} 3/3,~3' \
    --bind 'enter:become(nvim {1}  {2})'
}

which allows to "live-grep" (using here ripgrep) the files of jumper's database.